lsb's picture
add tokenizer
cc81c42
raw
history blame contribute delete
No virus
185 kB
{
"best_metric": 0.054694876074790955,
"best_model_checkpoint": "wav2vec2-base-allpemlsb-la/checkpoint-7500",
"epoch": 1.2853042602865894,
"global_step": 7500,
"is_hyper_param_search": false,
"is_local_process_zero": true,
"is_world_process_zero": true,
"log_history": [
{
"epoch": 0.0,
"learning_rate": 6.000000000000001e-08,
"loss": 21.313,
"step": 5
},
{
"epoch": 0.0,
"learning_rate": 1.6e-07,
"loss": 18.8113,
"step": 10
},
{
"epoch": 0.0,
"learning_rate": 2.6e-07,
"loss": 21.9003,
"step": 15
},
{
"epoch": 0.0,
"learning_rate": 3.6e-07,
"loss": 22.3353,
"step": 20
},
{
"epoch": 0.0,
"learning_rate": 4.6000000000000004e-07,
"loss": 22.3863,
"step": 25
},
{
"epoch": 0.01,
"learning_rate": 5.6e-07,
"loss": 22.2958,
"step": 30
},
{
"epoch": 0.01,
"learning_rate": 6.6e-07,
"loss": 22.3479,
"step": 35
},
{
"epoch": 0.01,
"learning_rate": 7.6e-07,
"loss": 22.3509,
"step": 40
},
{
"epoch": 0.01,
"learning_rate": 8.400000000000001e-07,
"loss": 22.1905,
"step": 45
},
{
"epoch": 0.01,
"learning_rate": 9.200000000000001e-07,
"loss": 23.6932,
"step": 50
},
{
"epoch": 0.01,
"learning_rate": 1.02e-06,
"loss": 17.4265,
"step": 55
},
{
"epoch": 0.01,
"learning_rate": 1.12e-06,
"loss": 17.0551,
"step": 60
},
{
"epoch": 0.01,
"learning_rate": 1.2200000000000002e-06,
"loss": 18.5504,
"step": 65
},
{
"epoch": 0.01,
"learning_rate": 1.32e-06,
"loss": 17.7868,
"step": 70
},
{
"epoch": 0.01,
"learning_rate": 1.42e-06,
"loss": 17.1873,
"step": 75
},
{
"epoch": 0.01,
"learning_rate": 1.52e-06,
"loss": 15.9765,
"step": 80
},
{
"epoch": 0.01,
"learning_rate": 1.6200000000000002e-06,
"loss": 15.3642,
"step": 85
},
{
"epoch": 0.02,
"learning_rate": 1.72e-06,
"loss": 13.7664,
"step": 90
},
{
"epoch": 0.02,
"learning_rate": 1.8200000000000002e-06,
"loss": 12.2769,
"step": 95
},
{
"epoch": 0.02,
"learning_rate": 1.9200000000000003e-06,
"loss": 12.2925,
"step": 100
},
{
"epoch": 0.02,
"learning_rate": 2.02e-06,
"loss": 8.818,
"step": 105
},
{
"epoch": 0.02,
"learning_rate": 2.12e-06,
"loss": 7.0128,
"step": 110
},
{
"epoch": 0.02,
"learning_rate": 2.2200000000000003e-06,
"loss": 7.2622,
"step": 115
},
{
"epoch": 0.02,
"learning_rate": 2.3200000000000002e-06,
"loss": 5.3161,
"step": 120
},
{
"epoch": 0.02,
"learning_rate": 2.42e-06,
"loss": 4.9828,
"step": 125
},
{
"epoch": 0.02,
"learning_rate": 2.52e-06,
"loss": 4.4695,
"step": 130
},
{
"epoch": 0.02,
"learning_rate": 2.6200000000000003e-06,
"loss": 4.664,
"step": 135
},
{
"epoch": 0.02,
"learning_rate": 2.7200000000000002e-06,
"loss": 4.0635,
"step": 140
},
{
"epoch": 0.02,
"learning_rate": 2.82e-06,
"loss": 4.4413,
"step": 145
},
{
"epoch": 0.03,
"learning_rate": 2.92e-06,
"loss": 4.2185,
"step": 150
},
{
"epoch": 0.03,
"learning_rate": 3.0200000000000003e-06,
"loss": 3.8167,
"step": 155
},
{
"epoch": 0.03,
"learning_rate": 3.12e-06,
"loss": 3.6831,
"step": 160
},
{
"epoch": 0.03,
"learning_rate": 3.2200000000000005e-06,
"loss": 3.4577,
"step": 165
},
{
"epoch": 0.03,
"learning_rate": 3.3200000000000004e-06,
"loss": 3.4685,
"step": 170
},
{
"epoch": 0.03,
"learning_rate": 3.4200000000000007e-06,
"loss": 3.4435,
"step": 175
},
{
"epoch": 0.03,
"learning_rate": 3.52e-06,
"loss": 3.2741,
"step": 180
},
{
"epoch": 0.03,
"learning_rate": 3.62e-06,
"loss": 3.7359,
"step": 185
},
{
"epoch": 0.03,
"learning_rate": 3.7200000000000004e-06,
"loss": 3.4724,
"step": 190
},
{
"epoch": 0.03,
"learning_rate": 3.820000000000001e-06,
"loss": 3.3526,
"step": 195
},
{
"epoch": 0.03,
"learning_rate": 3.920000000000001e-06,
"loss": 3.4805,
"step": 200
},
{
"epoch": 0.04,
"learning_rate": 4.0200000000000005e-06,
"loss": 3.2625,
"step": 205
},
{
"epoch": 0.04,
"learning_rate": 4.12e-06,
"loss": 3.1534,
"step": 210
},
{
"epoch": 0.04,
"learning_rate": 4.22e-06,
"loss": 3.207,
"step": 215
},
{
"epoch": 0.04,
"learning_rate": 4.32e-06,
"loss": 3.1436,
"step": 220
},
{
"epoch": 0.04,
"learning_rate": 4.42e-06,
"loss": 3.0799,
"step": 225
},
{
"epoch": 0.04,
"learning_rate": 4.520000000000001e-06,
"loss": 3.233,
"step": 230
},
{
"epoch": 0.04,
"learning_rate": 4.620000000000001e-06,
"loss": 3.1099,
"step": 235
},
{
"epoch": 0.04,
"learning_rate": 4.7200000000000005e-06,
"loss": 3.1029,
"step": 240
},
{
"epoch": 0.04,
"learning_rate": 4.8200000000000004e-06,
"loss": 3.128,
"step": 245
},
{
"epoch": 0.04,
"learning_rate": 4.92e-06,
"loss": 3.1941,
"step": 250
},
{
"epoch": 0.04,
"learning_rate": 5.02e-06,
"loss": 3.2538,
"step": 255
},
{
"epoch": 0.04,
"learning_rate": 5.12e-06,
"loss": 3.0986,
"step": 260
},
{
"epoch": 0.05,
"learning_rate": 5.220000000000001e-06,
"loss": 3.0339,
"step": 265
},
{
"epoch": 0.05,
"learning_rate": 5.320000000000001e-06,
"loss": 3.0984,
"step": 270
},
{
"epoch": 0.05,
"learning_rate": 5.420000000000001e-06,
"loss": 3.1802,
"step": 275
},
{
"epoch": 0.05,
"learning_rate": 5.5200000000000005e-06,
"loss": 3.071,
"step": 280
},
{
"epoch": 0.05,
"learning_rate": 5.620000000000001e-06,
"loss": 3.0324,
"step": 285
},
{
"epoch": 0.05,
"learning_rate": 5.72e-06,
"loss": 3.1339,
"step": 290
},
{
"epoch": 0.05,
"learning_rate": 5.82e-06,
"loss": 3.1202,
"step": 295
},
{
"epoch": 0.05,
"learning_rate": 5.92e-06,
"loss": 3.1724,
"step": 300
},
{
"epoch": 0.05,
"learning_rate": 6.02e-06,
"loss": 3.2735,
"step": 305
},
{
"epoch": 0.05,
"learning_rate": 6.120000000000001e-06,
"loss": 3.0724,
"step": 310
},
{
"epoch": 0.05,
"learning_rate": 6.220000000000001e-06,
"loss": 3.0353,
"step": 315
},
{
"epoch": 0.05,
"learning_rate": 6.3200000000000005e-06,
"loss": 3.0697,
"step": 320
},
{
"epoch": 0.06,
"learning_rate": 6.42e-06,
"loss": 3.0404,
"step": 325
},
{
"epoch": 0.06,
"learning_rate": 6.520000000000001e-06,
"loss": 3.0535,
"step": 330
},
{
"epoch": 0.06,
"learning_rate": 6.620000000000001e-06,
"loss": 3.0306,
"step": 335
},
{
"epoch": 0.06,
"learning_rate": 6.720000000000001e-06,
"loss": 3.0583,
"step": 340
},
{
"epoch": 0.06,
"learning_rate": 6.820000000000001e-06,
"loss": 3.0224,
"step": 345
},
{
"epoch": 0.06,
"learning_rate": 6.92e-06,
"loss": 3.1336,
"step": 350
},
{
"epoch": 0.06,
"learning_rate": 7.0200000000000006e-06,
"loss": 3.1767,
"step": 355
},
{
"epoch": 0.06,
"learning_rate": 7.1200000000000004e-06,
"loss": 3.0263,
"step": 360
},
{
"epoch": 0.06,
"learning_rate": 7.22e-06,
"loss": 2.99,
"step": 365
},
{
"epoch": 0.06,
"learning_rate": 7.32e-06,
"loss": 2.9538,
"step": 370
},
{
"epoch": 0.06,
"learning_rate": 7.420000000000001e-06,
"loss": 2.9597,
"step": 375
},
{
"epoch": 0.07,
"learning_rate": 7.520000000000001e-06,
"loss": 2.9824,
"step": 380
},
{
"epoch": 0.07,
"learning_rate": 7.620000000000001e-06,
"loss": 3.0408,
"step": 385
},
{
"epoch": 0.07,
"learning_rate": 7.72e-06,
"loss": 2.9733,
"step": 390
},
{
"epoch": 0.07,
"learning_rate": 7.820000000000001e-06,
"loss": 3.0239,
"step": 395
},
{
"epoch": 0.07,
"learning_rate": 7.92e-06,
"loss": 3.0879,
"step": 400
},
{
"epoch": 0.07,
"learning_rate": 8.020000000000001e-06,
"loss": 3.1678,
"step": 405
},
{
"epoch": 0.07,
"learning_rate": 8.120000000000002e-06,
"loss": 3.0097,
"step": 410
},
{
"epoch": 0.07,
"learning_rate": 8.220000000000001e-06,
"loss": 2.915,
"step": 415
},
{
"epoch": 0.07,
"learning_rate": 8.32e-06,
"loss": 2.9152,
"step": 420
},
{
"epoch": 0.07,
"learning_rate": 8.42e-06,
"loss": 2.8817,
"step": 425
},
{
"epoch": 0.07,
"learning_rate": 8.52e-06,
"loss": 2.8852,
"step": 430
},
{
"epoch": 0.07,
"learning_rate": 8.62e-06,
"loss": 2.9103,
"step": 435
},
{
"epoch": 0.08,
"learning_rate": 8.720000000000001e-06,
"loss": 2.8642,
"step": 440
},
{
"epoch": 0.08,
"learning_rate": 8.82e-06,
"loss": 2.9038,
"step": 445
},
{
"epoch": 0.08,
"learning_rate": 8.920000000000001e-06,
"loss": 2.9846,
"step": 450
},
{
"epoch": 0.08,
"learning_rate": 9.020000000000002e-06,
"loss": 3.0014,
"step": 455
},
{
"epoch": 0.08,
"learning_rate": 9.12e-06,
"loss": 2.7939,
"step": 460
},
{
"epoch": 0.08,
"learning_rate": 9.220000000000002e-06,
"loss": 2.6869,
"step": 465
},
{
"epoch": 0.08,
"learning_rate": 9.32e-06,
"loss": 2.7217,
"step": 470
},
{
"epoch": 0.08,
"learning_rate": 9.42e-06,
"loss": 2.7731,
"step": 475
},
{
"epoch": 0.08,
"learning_rate": 9.52e-06,
"loss": 2.6662,
"step": 480
},
{
"epoch": 0.08,
"learning_rate": 9.620000000000001e-06,
"loss": 2.6541,
"step": 485
},
{
"epoch": 0.08,
"learning_rate": 9.72e-06,
"loss": 2.5936,
"step": 490
},
{
"epoch": 0.08,
"learning_rate": 9.820000000000001e-06,
"loss": 2.6427,
"step": 495
},
{
"epoch": 0.09,
"learning_rate": 9.920000000000002e-06,
"loss": 2.7209,
"step": 500
},
{
"epoch": 0.09,
"eval_loss": 2.4995110034942627,
"eval_runtime": 1351.4853,
"eval_samples_per_second": 15.353,
"eval_steps_per_second": 3.839,
"eval_wer": 1.0,
"step": 500
},
{
"epoch": 0.09,
"learning_rate": 1.002e-05,
"loss": 2.7397,
"step": 505
},
{
"epoch": 0.09,
"learning_rate": 1.0120000000000001e-05,
"loss": 2.4777,
"step": 510
},
{
"epoch": 0.09,
"learning_rate": 1.022e-05,
"loss": 2.4546,
"step": 515
},
{
"epoch": 0.09,
"learning_rate": 1.0320000000000001e-05,
"loss": 2.4474,
"step": 520
},
{
"epoch": 0.09,
"learning_rate": 1.0420000000000002e-05,
"loss": 2.3956,
"step": 525
},
{
"epoch": 0.09,
"learning_rate": 1.0520000000000001e-05,
"loss": 2.3162,
"step": 530
},
{
"epoch": 0.09,
"learning_rate": 1.0620000000000002e-05,
"loss": 2.3583,
"step": 535
},
{
"epoch": 0.09,
"learning_rate": 1.072e-05,
"loss": 2.3472,
"step": 540
},
{
"epoch": 0.09,
"learning_rate": 1.0820000000000001e-05,
"loss": 2.2247,
"step": 545
},
{
"epoch": 0.09,
"learning_rate": 1.0920000000000002e-05,
"loss": 2.3901,
"step": 550
},
{
"epoch": 0.1,
"learning_rate": 1.1020000000000001e-05,
"loss": 2.4117,
"step": 555
},
{
"epoch": 0.1,
"learning_rate": 1.1120000000000002e-05,
"loss": 2.1732,
"step": 560
},
{
"epoch": 0.1,
"learning_rate": 1.1220000000000003e-05,
"loss": 2.0335,
"step": 565
},
{
"epoch": 0.1,
"learning_rate": 1.132e-05,
"loss": 2.1071,
"step": 570
},
{
"epoch": 0.1,
"learning_rate": 1.142e-05,
"loss": 2.0072,
"step": 575
},
{
"epoch": 0.1,
"learning_rate": 1.152e-05,
"loss": 2.0084,
"step": 580
},
{
"epoch": 0.1,
"learning_rate": 1.162e-05,
"loss": 1.99,
"step": 585
},
{
"epoch": 0.1,
"learning_rate": 1.172e-05,
"loss": 1.9867,
"step": 590
},
{
"epoch": 0.1,
"learning_rate": 1.182e-05,
"loss": 1.9144,
"step": 595
},
{
"epoch": 0.1,
"learning_rate": 1.1920000000000001e-05,
"loss": 2.2666,
"step": 600
},
{
"epoch": 0.1,
"learning_rate": 1.202e-05,
"loss": 2.1645,
"step": 605
},
{
"epoch": 0.1,
"learning_rate": 1.2120000000000001e-05,
"loss": 1.9345,
"step": 610
},
{
"epoch": 0.11,
"learning_rate": 1.2220000000000002e-05,
"loss": 1.894,
"step": 615
},
{
"epoch": 0.11,
"learning_rate": 1.232e-05,
"loss": 1.7511,
"step": 620
},
{
"epoch": 0.11,
"learning_rate": 1.2420000000000001e-05,
"loss": 1.7202,
"step": 625
},
{
"epoch": 0.11,
"learning_rate": 1.252e-05,
"loss": 1.6494,
"step": 630
},
{
"epoch": 0.11,
"learning_rate": 1.2620000000000001e-05,
"loss": 1.5629,
"step": 635
},
{
"epoch": 0.11,
"learning_rate": 1.2720000000000002e-05,
"loss": 1.706,
"step": 640
},
{
"epoch": 0.11,
"learning_rate": 1.2820000000000001e-05,
"loss": 1.7459,
"step": 645
},
{
"epoch": 0.11,
"learning_rate": 1.2920000000000002e-05,
"loss": 1.8798,
"step": 650
},
{
"epoch": 0.11,
"learning_rate": 1.302e-05,
"loss": 1.9824,
"step": 655
},
{
"epoch": 0.11,
"learning_rate": 1.3120000000000001e-05,
"loss": 1.6461,
"step": 660
},
{
"epoch": 0.11,
"learning_rate": 1.3220000000000002e-05,
"loss": 1.3982,
"step": 665
},
{
"epoch": 0.11,
"learning_rate": 1.3320000000000001e-05,
"loss": 1.4166,
"step": 670
},
{
"epoch": 0.12,
"learning_rate": 1.3420000000000002e-05,
"loss": 1.3656,
"step": 675
},
{
"epoch": 0.12,
"learning_rate": 1.3520000000000003e-05,
"loss": 1.4433,
"step": 680
},
{
"epoch": 0.12,
"learning_rate": 1.3620000000000002e-05,
"loss": 1.5055,
"step": 685
},
{
"epoch": 0.12,
"learning_rate": 1.3720000000000002e-05,
"loss": 1.4942,
"step": 690
},
{
"epoch": 0.12,
"learning_rate": 1.382e-05,
"loss": 1.4016,
"step": 695
},
{
"epoch": 0.12,
"learning_rate": 1.392e-05,
"loss": 1.5213,
"step": 700
},
{
"epoch": 0.12,
"learning_rate": 1.402e-05,
"loss": 1.6288,
"step": 705
},
{
"epoch": 0.12,
"learning_rate": 1.412e-05,
"loss": 1.2364,
"step": 710
},
{
"epoch": 0.12,
"learning_rate": 1.4220000000000001e-05,
"loss": 1.0673,
"step": 715
},
{
"epoch": 0.12,
"learning_rate": 1.432e-05,
"loss": 1.0581,
"step": 720
},
{
"epoch": 0.12,
"learning_rate": 1.4420000000000001e-05,
"loss": 1.1221,
"step": 725
},
{
"epoch": 0.13,
"learning_rate": 1.4520000000000002e-05,
"loss": 0.9885,
"step": 730
},
{
"epoch": 0.13,
"learning_rate": 1.462e-05,
"loss": 1.0691,
"step": 735
},
{
"epoch": 0.13,
"learning_rate": 1.4720000000000001e-05,
"loss": 1.0709,
"step": 740
},
{
"epoch": 0.13,
"learning_rate": 1.482e-05,
"loss": 1.2689,
"step": 745
},
{
"epoch": 0.13,
"learning_rate": 1.4920000000000001e-05,
"loss": 1.4491,
"step": 750
},
{
"epoch": 0.13,
"learning_rate": 1.5020000000000002e-05,
"loss": 1.556,
"step": 755
},
{
"epoch": 0.13,
"learning_rate": 1.5120000000000001e-05,
"loss": 1.1487,
"step": 760
},
{
"epoch": 0.13,
"learning_rate": 1.5220000000000002e-05,
"loss": 0.9499,
"step": 765
},
{
"epoch": 0.13,
"learning_rate": 1.5320000000000002e-05,
"loss": 0.8911,
"step": 770
},
{
"epoch": 0.13,
"learning_rate": 1.542e-05,
"loss": 1.0133,
"step": 775
},
{
"epoch": 0.13,
"learning_rate": 1.552e-05,
"loss": 0.988,
"step": 780
},
{
"epoch": 0.13,
"learning_rate": 1.5620000000000003e-05,
"loss": 0.7904,
"step": 785
},
{
"epoch": 0.14,
"learning_rate": 1.5720000000000002e-05,
"loss": 0.9463,
"step": 790
},
{
"epoch": 0.14,
"learning_rate": 1.582e-05,
"loss": 1.0314,
"step": 795
},
{
"epoch": 0.14,
"learning_rate": 1.5920000000000003e-05,
"loss": 1.3329,
"step": 800
},
{
"epoch": 0.14,
"learning_rate": 1.6020000000000002e-05,
"loss": 1.4135,
"step": 805
},
{
"epoch": 0.14,
"learning_rate": 1.612e-05,
"loss": 1.0044,
"step": 810
},
{
"epoch": 0.14,
"learning_rate": 1.6220000000000004e-05,
"loss": 0.7801,
"step": 815
},
{
"epoch": 0.14,
"learning_rate": 1.632e-05,
"loss": 0.8437,
"step": 820
},
{
"epoch": 0.14,
"learning_rate": 1.6420000000000002e-05,
"loss": 0.8899,
"step": 825
},
{
"epoch": 0.14,
"learning_rate": 1.652e-05,
"loss": 0.9004,
"step": 830
},
{
"epoch": 0.14,
"learning_rate": 1.662e-05,
"loss": 0.8363,
"step": 835
},
{
"epoch": 0.14,
"learning_rate": 1.672e-05,
"loss": 0.8766,
"step": 840
},
{
"epoch": 0.14,
"learning_rate": 1.682e-05,
"loss": 1.0172,
"step": 845
},
{
"epoch": 0.15,
"learning_rate": 1.692e-05,
"loss": 1.2288,
"step": 850
},
{
"epoch": 0.15,
"learning_rate": 1.702e-05,
"loss": 1.3441,
"step": 855
},
{
"epoch": 0.15,
"learning_rate": 1.7120000000000002e-05,
"loss": 0.8382,
"step": 860
},
{
"epoch": 0.15,
"learning_rate": 1.722e-05,
"loss": 0.5626,
"step": 865
},
{
"epoch": 0.15,
"learning_rate": 1.732e-05,
"loss": 0.7224,
"step": 870
},
{
"epoch": 0.15,
"learning_rate": 1.7420000000000003e-05,
"loss": 0.751,
"step": 875
},
{
"epoch": 0.15,
"learning_rate": 1.752e-05,
"loss": 0.7194,
"step": 880
},
{
"epoch": 0.15,
"learning_rate": 1.762e-05,
"loss": 0.8238,
"step": 885
},
{
"epoch": 0.15,
"learning_rate": 1.7720000000000003e-05,
"loss": 0.8902,
"step": 890
},
{
"epoch": 0.15,
"learning_rate": 1.7820000000000002e-05,
"loss": 0.9449,
"step": 895
},
{
"epoch": 0.15,
"learning_rate": 1.792e-05,
"loss": 1.1581,
"step": 900
},
{
"epoch": 0.16,
"learning_rate": 1.802e-05,
"loss": 1.1919,
"step": 905
},
{
"epoch": 0.16,
"learning_rate": 1.8120000000000003e-05,
"loss": 0.87,
"step": 910
},
{
"epoch": 0.16,
"learning_rate": 1.8220000000000002e-05,
"loss": 0.6561,
"step": 915
},
{
"epoch": 0.16,
"learning_rate": 1.832e-05,
"loss": 0.7263,
"step": 920
},
{
"epoch": 0.16,
"learning_rate": 1.8420000000000003e-05,
"loss": 0.7124,
"step": 925
},
{
"epoch": 0.16,
"learning_rate": 1.8520000000000002e-05,
"loss": 0.7209,
"step": 930
},
{
"epoch": 0.16,
"learning_rate": 1.862e-05,
"loss": 0.7385,
"step": 935
},
{
"epoch": 0.16,
"learning_rate": 1.8720000000000004e-05,
"loss": 0.7236,
"step": 940
},
{
"epoch": 0.16,
"learning_rate": 1.882e-05,
"loss": 1.008,
"step": 945
},
{
"epoch": 0.16,
"learning_rate": 1.8920000000000002e-05,
"loss": 1.0779,
"step": 950
},
{
"epoch": 0.16,
"learning_rate": 1.902e-05,
"loss": 1.1714,
"step": 955
},
{
"epoch": 0.16,
"learning_rate": 1.912e-05,
"loss": 0.8129,
"step": 960
},
{
"epoch": 0.17,
"learning_rate": 1.9220000000000002e-05,
"loss": 0.6949,
"step": 965
},
{
"epoch": 0.17,
"learning_rate": 1.932e-05,
"loss": 0.6265,
"step": 970
},
{
"epoch": 0.17,
"learning_rate": 1.942e-05,
"loss": 0.5718,
"step": 975
},
{
"epoch": 0.17,
"learning_rate": 1.9520000000000003e-05,
"loss": 0.7736,
"step": 980
},
{
"epoch": 0.17,
"learning_rate": 1.9620000000000002e-05,
"loss": 0.7621,
"step": 985
},
{
"epoch": 0.17,
"learning_rate": 1.972e-05,
"loss": 0.7196,
"step": 990
},
{
"epoch": 0.17,
"learning_rate": 1.982e-05,
"loss": 0.9488,
"step": 995
},
{
"epoch": 0.17,
"learning_rate": 1.9920000000000002e-05,
"loss": 1.0392,
"step": 1000
},
{
"epoch": 0.17,
"eval_loss": 0.3022914528846741,
"eval_runtime": 1330.6014,
"eval_samples_per_second": 15.594,
"eval_steps_per_second": 3.899,
"eval_wer": 0.6319072606553866,
"step": 1000
},
{
"epoch": 0.17,
"learning_rate": 1.9999562171628723e-05,
"loss": 1.1876,
"step": 1005
},
{
"epoch": 0.17,
"learning_rate": 1.9997373029772333e-05,
"loss": 0.7495,
"step": 1010
},
{
"epoch": 0.17,
"learning_rate": 1.999518388791594e-05,
"loss": 0.6574,
"step": 1015
},
{
"epoch": 0.17,
"learning_rate": 1.9992994746059546e-05,
"loss": 0.5601,
"step": 1020
},
{
"epoch": 0.18,
"learning_rate": 1.9990805604203155e-05,
"loss": 0.5899,
"step": 1025
},
{
"epoch": 0.18,
"learning_rate": 1.9988616462346762e-05,
"loss": 0.5952,
"step": 1030
},
{
"epoch": 0.18,
"learning_rate": 1.998642732049037e-05,
"loss": 0.6568,
"step": 1035
},
{
"epoch": 0.18,
"learning_rate": 1.9984238178633978e-05,
"loss": 0.7713,
"step": 1040
},
{
"epoch": 0.18,
"learning_rate": 1.9982049036777585e-05,
"loss": 0.8182,
"step": 1045
},
{
"epoch": 0.18,
"learning_rate": 1.997985989492119e-05,
"loss": 0.9574,
"step": 1050
},
{
"epoch": 0.18,
"learning_rate": 1.99776707530648e-05,
"loss": 1.0197,
"step": 1055
},
{
"epoch": 0.18,
"learning_rate": 1.9975481611208407e-05,
"loss": 0.6908,
"step": 1060
},
{
"epoch": 0.18,
"learning_rate": 1.9973292469352014e-05,
"loss": 0.5128,
"step": 1065
},
{
"epoch": 0.18,
"learning_rate": 1.9971103327495624e-05,
"loss": 0.5686,
"step": 1070
},
{
"epoch": 0.18,
"learning_rate": 1.9968914185639233e-05,
"loss": 0.5416,
"step": 1075
},
{
"epoch": 0.19,
"learning_rate": 1.996672504378284e-05,
"loss": 0.5784,
"step": 1080
},
{
"epoch": 0.19,
"learning_rate": 1.9964535901926446e-05,
"loss": 0.5869,
"step": 1085
},
{
"epoch": 0.19,
"learning_rate": 1.9962346760070056e-05,
"loss": 0.6784,
"step": 1090
},
{
"epoch": 0.19,
"learning_rate": 1.9960157618213662e-05,
"loss": 0.7512,
"step": 1095
},
{
"epoch": 0.19,
"learning_rate": 1.995796847635727e-05,
"loss": 1.0341,
"step": 1100
},
{
"epoch": 0.19,
"learning_rate": 1.995577933450088e-05,
"loss": 0.9844,
"step": 1105
},
{
"epoch": 0.19,
"learning_rate": 1.9953590192644485e-05,
"loss": 0.6568,
"step": 1110
},
{
"epoch": 0.19,
"learning_rate": 1.995140105078809e-05,
"loss": 0.4461,
"step": 1115
},
{
"epoch": 0.19,
"learning_rate": 1.99492119089317e-05,
"loss": 0.4822,
"step": 1120
},
{
"epoch": 0.19,
"learning_rate": 1.9947022767075308e-05,
"loss": 0.456,
"step": 1125
},
{
"epoch": 0.19,
"learning_rate": 1.9944833625218914e-05,
"loss": 0.5295,
"step": 1130
},
{
"epoch": 0.19,
"learning_rate": 1.9942644483362524e-05,
"loss": 0.618,
"step": 1135
},
{
"epoch": 0.2,
"learning_rate": 1.994045534150613e-05,
"loss": 0.664,
"step": 1140
},
{
"epoch": 0.2,
"learning_rate": 1.9938266199649737e-05,
"loss": 0.6447,
"step": 1145
},
{
"epoch": 0.2,
"learning_rate": 1.9936077057793347e-05,
"loss": 0.9506,
"step": 1150
},
{
"epoch": 0.2,
"learning_rate": 1.9933887915936953e-05,
"loss": 1.0024,
"step": 1155
},
{
"epoch": 0.2,
"learning_rate": 1.993169877408056e-05,
"loss": 0.6365,
"step": 1160
},
{
"epoch": 0.2,
"learning_rate": 1.992950963222417e-05,
"loss": 0.4754,
"step": 1165
},
{
"epoch": 0.2,
"learning_rate": 1.9927320490367776e-05,
"loss": 0.4488,
"step": 1170
},
{
"epoch": 0.2,
"learning_rate": 1.9925131348511386e-05,
"loss": 0.6566,
"step": 1175
},
{
"epoch": 0.2,
"learning_rate": 1.9922942206654992e-05,
"loss": 0.5274,
"step": 1180
},
{
"epoch": 0.2,
"learning_rate": 1.9920753064798602e-05,
"loss": 0.5331,
"step": 1185
},
{
"epoch": 0.2,
"learning_rate": 1.991856392294221e-05,
"loss": 0.5909,
"step": 1190
},
{
"epoch": 0.2,
"learning_rate": 1.9916374781085815e-05,
"loss": 0.6535,
"step": 1195
},
{
"epoch": 0.21,
"learning_rate": 1.9914185639229425e-05,
"loss": 0.8616,
"step": 1200
},
{
"epoch": 0.21,
"learning_rate": 1.991199649737303e-05,
"loss": 0.8683,
"step": 1205
},
{
"epoch": 0.21,
"learning_rate": 1.990980735551664e-05,
"loss": 0.6498,
"step": 1210
},
{
"epoch": 0.21,
"learning_rate": 1.9907618213660247e-05,
"loss": 0.4443,
"step": 1215
},
{
"epoch": 0.21,
"learning_rate": 1.9905429071803854e-05,
"loss": 0.362,
"step": 1220
},
{
"epoch": 0.21,
"learning_rate": 1.9903239929947464e-05,
"loss": 0.4395,
"step": 1225
},
{
"epoch": 0.21,
"learning_rate": 1.990105078809107e-05,
"loss": 0.5171,
"step": 1230
},
{
"epoch": 0.21,
"learning_rate": 1.9898861646234677e-05,
"loss": 0.5486,
"step": 1235
},
{
"epoch": 0.21,
"learning_rate": 1.9896672504378286e-05,
"loss": 0.5423,
"step": 1240
},
{
"epoch": 0.21,
"learning_rate": 1.9894483362521893e-05,
"loss": 0.7308,
"step": 1245
},
{
"epoch": 0.21,
"learning_rate": 1.98922942206655e-05,
"loss": 0.8619,
"step": 1250
},
{
"epoch": 0.22,
"learning_rate": 1.989010507880911e-05,
"loss": 0.867,
"step": 1255
},
{
"epoch": 0.22,
"learning_rate": 1.9887915936952716e-05,
"loss": 0.5763,
"step": 1260
},
{
"epoch": 0.22,
"learning_rate": 1.9885726795096322e-05,
"loss": 0.3397,
"step": 1265
},
{
"epoch": 0.22,
"learning_rate": 1.9883537653239932e-05,
"loss": 0.456,
"step": 1270
},
{
"epoch": 0.22,
"learning_rate": 1.988134851138354e-05,
"loss": 0.4711,
"step": 1275
},
{
"epoch": 0.22,
"learning_rate": 1.9879159369527148e-05,
"loss": 0.4648,
"step": 1280
},
{
"epoch": 0.22,
"learning_rate": 1.9876970227670754e-05,
"loss": 0.4536,
"step": 1285
},
{
"epoch": 0.22,
"learning_rate": 1.9874781085814364e-05,
"loss": 0.4834,
"step": 1290
},
{
"epoch": 0.22,
"learning_rate": 1.987259194395797e-05,
"loss": 0.7299,
"step": 1295
},
{
"epoch": 0.22,
"learning_rate": 1.9870402802101577e-05,
"loss": 0.8013,
"step": 1300
},
{
"epoch": 0.22,
"learning_rate": 1.9868213660245187e-05,
"loss": 0.8243,
"step": 1305
},
{
"epoch": 0.22,
"learning_rate": 1.9866024518388793e-05,
"loss": 0.5073,
"step": 1310
},
{
"epoch": 0.23,
"learning_rate": 1.98638353765324e-05,
"loss": 0.3678,
"step": 1315
},
{
"epoch": 0.23,
"learning_rate": 1.986164623467601e-05,
"loss": 0.3777,
"step": 1320
},
{
"epoch": 0.23,
"learning_rate": 1.9859457092819616e-05,
"loss": 0.471,
"step": 1325
},
{
"epoch": 0.23,
"learning_rate": 1.9857267950963223e-05,
"loss": 0.424,
"step": 1330
},
{
"epoch": 0.23,
"learning_rate": 1.9855078809106832e-05,
"loss": 0.5631,
"step": 1335
},
{
"epoch": 0.23,
"learning_rate": 1.985288966725044e-05,
"loss": 0.5752,
"step": 1340
},
{
"epoch": 0.23,
"learning_rate": 1.9850700525394045e-05,
"loss": 0.6967,
"step": 1345
},
{
"epoch": 0.23,
"learning_rate": 1.9848511383537655e-05,
"loss": 0.8207,
"step": 1350
},
{
"epoch": 0.23,
"learning_rate": 1.984632224168126e-05,
"loss": 0.7361,
"step": 1355
},
{
"epoch": 0.23,
"learning_rate": 1.9844133099824868e-05,
"loss": 0.4835,
"step": 1360
},
{
"epoch": 0.23,
"learning_rate": 1.9841943957968478e-05,
"loss": 0.3658,
"step": 1365
},
{
"epoch": 0.23,
"learning_rate": 1.9839754816112088e-05,
"loss": 0.3357,
"step": 1370
},
{
"epoch": 0.24,
"learning_rate": 1.9837565674255694e-05,
"loss": 0.3789,
"step": 1375
},
{
"epoch": 0.24,
"learning_rate": 1.98353765323993e-05,
"loss": 0.3966,
"step": 1380
},
{
"epoch": 0.24,
"learning_rate": 1.983318739054291e-05,
"loss": 0.4599,
"step": 1385
},
{
"epoch": 0.24,
"learning_rate": 1.9830998248686517e-05,
"loss": 0.4357,
"step": 1390
},
{
"epoch": 0.24,
"learning_rate": 1.9828809106830123e-05,
"loss": 0.6543,
"step": 1395
},
{
"epoch": 0.24,
"learning_rate": 1.9826619964973733e-05,
"loss": 0.8527,
"step": 1400
},
{
"epoch": 0.24,
"learning_rate": 1.982443082311734e-05,
"loss": 0.7114,
"step": 1405
},
{
"epoch": 0.24,
"learning_rate": 1.9822241681260946e-05,
"loss": 0.4986,
"step": 1410
},
{
"epoch": 0.24,
"learning_rate": 1.9820052539404556e-05,
"loss": 0.3365,
"step": 1415
},
{
"epoch": 0.24,
"learning_rate": 1.9817863397548162e-05,
"loss": 0.3406,
"step": 1420
},
{
"epoch": 0.24,
"learning_rate": 1.981567425569177e-05,
"loss": 0.4034,
"step": 1425
},
{
"epoch": 0.25,
"learning_rate": 1.981348511383538e-05,
"loss": 0.4175,
"step": 1430
},
{
"epoch": 0.25,
"learning_rate": 1.9811295971978985e-05,
"loss": 0.4336,
"step": 1435
},
{
"epoch": 0.25,
"learning_rate": 1.9809106830122595e-05,
"loss": 0.4741,
"step": 1440
},
{
"epoch": 0.25,
"learning_rate": 1.98069176882662e-05,
"loss": 0.5576,
"step": 1445
},
{
"epoch": 0.25,
"learning_rate": 1.9804728546409808e-05,
"loss": 0.7395,
"step": 1450
},
{
"epoch": 0.25,
"learning_rate": 1.9802539404553417e-05,
"loss": 0.7375,
"step": 1455
},
{
"epoch": 0.25,
"learning_rate": 1.9800350262697024e-05,
"loss": 0.4683,
"step": 1460
},
{
"epoch": 0.25,
"learning_rate": 1.979816112084063e-05,
"loss": 0.2967,
"step": 1465
},
{
"epoch": 0.25,
"learning_rate": 1.979597197898424e-05,
"loss": 0.3669,
"step": 1470
},
{
"epoch": 0.25,
"learning_rate": 1.979378283712785e-05,
"loss": 0.3353,
"step": 1475
},
{
"epoch": 0.25,
"learning_rate": 1.9791593695271456e-05,
"loss": 0.3805,
"step": 1480
},
{
"epoch": 0.25,
"learning_rate": 1.9789404553415063e-05,
"loss": 0.4116,
"step": 1485
},
{
"epoch": 0.26,
"learning_rate": 1.9787215411558673e-05,
"loss": 0.4539,
"step": 1490
},
{
"epoch": 0.26,
"learning_rate": 1.978502626970228e-05,
"loss": 0.5322,
"step": 1495
},
{
"epoch": 0.26,
"learning_rate": 1.9782837127845885e-05,
"loss": 0.7897,
"step": 1500
},
{
"epoch": 0.26,
"eval_loss": 0.21090660989284515,
"eval_runtime": 1309.9451,
"eval_samples_per_second": 15.84,
"eval_steps_per_second": 3.96,
"eval_wer": 0.5727404155065324,
"step": 1500
},
{
"epoch": 0.26,
"learning_rate": 1.9780647985989495e-05,
"loss": 0.6925,
"step": 1505
},
{
"epoch": 0.26,
"learning_rate": 1.9778458844133102e-05,
"loss": 0.4558,
"step": 1510
},
{
"epoch": 0.26,
"learning_rate": 1.9776269702276708e-05,
"loss": 0.2934,
"step": 1515
},
{
"epoch": 0.26,
"learning_rate": 1.9774080560420318e-05,
"loss": 0.3337,
"step": 1520
},
{
"epoch": 0.26,
"learning_rate": 1.9771891418563924e-05,
"loss": 0.3068,
"step": 1525
},
{
"epoch": 0.26,
"learning_rate": 1.976970227670753e-05,
"loss": 0.3689,
"step": 1530
},
{
"epoch": 0.26,
"learning_rate": 1.976751313485114e-05,
"loss": 0.4243,
"step": 1535
},
{
"epoch": 0.26,
"learning_rate": 1.9765323992994747e-05,
"loss": 0.4972,
"step": 1540
},
{
"epoch": 0.26,
"learning_rate": 1.9763134851138354e-05,
"loss": 0.4774,
"step": 1545
},
{
"epoch": 0.27,
"learning_rate": 1.9760945709281963e-05,
"loss": 0.7317,
"step": 1550
},
{
"epoch": 0.27,
"learning_rate": 1.975875656742557e-05,
"loss": 0.662,
"step": 1555
},
{
"epoch": 0.27,
"learning_rate": 1.9756567425569176e-05,
"loss": 0.3663,
"step": 1560
},
{
"epoch": 0.27,
"learning_rate": 1.9754378283712786e-05,
"loss": 0.2623,
"step": 1565
},
{
"epoch": 0.27,
"learning_rate": 1.9752189141856396e-05,
"loss": 0.3906,
"step": 1570
},
{
"epoch": 0.27,
"learning_rate": 1.9750000000000002e-05,
"loss": 0.315,
"step": 1575
},
{
"epoch": 0.27,
"learning_rate": 1.974781085814361e-05,
"loss": 0.3034,
"step": 1580
},
{
"epoch": 0.27,
"learning_rate": 1.974562171628722e-05,
"loss": 0.4665,
"step": 1585
},
{
"epoch": 0.27,
"learning_rate": 1.9743432574430825e-05,
"loss": 0.4834,
"step": 1590
},
{
"epoch": 0.27,
"learning_rate": 1.974124343257443e-05,
"loss": 0.5543,
"step": 1595
},
{
"epoch": 0.27,
"learning_rate": 1.973905429071804e-05,
"loss": 0.7425,
"step": 1600
},
{
"epoch": 0.28,
"learning_rate": 1.9736865148861648e-05,
"loss": 0.6905,
"step": 1605
},
{
"epoch": 0.28,
"learning_rate": 1.9734676007005254e-05,
"loss": 0.42,
"step": 1610
},
{
"epoch": 0.28,
"learning_rate": 1.9732486865148864e-05,
"loss": 0.263,
"step": 1615
},
{
"epoch": 0.28,
"learning_rate": 1.973029772329247e-05,
"loss": 0.2688,
"step": 1620
},
{
"epoch": 0.28,
"learning_rate": 1.9728108581436077e-05,
"loss": 0.32,
"step": 1625
},
{
"epoch": 0.28,
"learning_rate": 1.9725919439579687e-05,
"loss": 0.4025,
"step": 1630
},
{
"epoch": 0.28,
"learning_rate": 1.9723730297723293e-05,
"loss": 0.4189,
"step": 1635
},
{
"epoch": 0.28,
"learning_rate": 1.97215411558669e-05,
"loss": 0.437,
"step": 1640
},
{
"epoch": 0.28,
"learning_rate": 1.971935201401051e-05,
"loss": 0.5859,
"step": 1645
},
{
"epoch": 0.28,
"learning_rate": 1.9717162872154116e-05,
"loss": 0.7288,
"step": 1650
},
{
"epoch": 0.28,
"learning_rate": 1.9714973730297726e-05,
"loss": 0.6683,
"step": 1655
},
{
"epoch": 0.28,
"learning_rate": 1.9712784588441332e-05,
"loss": 0.505,
"step": 1660
},
{
"epoch": 0.29,
"learning_rate": 1.9710595446584942e-05,
"loss": 0.2271,
"step": 1665
},
{
"epoch": 0.29,
"learning_rate": 1.970840630472855e-05,
"loss": 0.273,
"step": 1670
},
{
"epoch": 0.29,
"learning_rate": 1.9706217162872158e-05,
"loss": 0.3114,
"step": 1675
},
{
"epoch": 0.29,
"learning_rate": 1.9704028021015765e-05,
"loss": 0.3687,
"step": 1680
},
{
"epoch": 0.29,
"learning_rate": 1.970183887915937e-05,
"loss": 0.3806,
"step": 1685
},
{
"epoch": 0.29,
"learning_rate": 1.969964973730298e-05,
"loss": 0.4416,
"step": 1690
},
{
"epoch": 0.29,
"learning_rate": 1.9697460595446587e-05,
"loss": 0.5305,
"step": 1695
},
{
"epoch": 0.29,
"learning_rate": 1.9695271453590194e-05,
"loss": 0.799,
"step": 1700
},
{
"epoch": 0.29,
"learning_rate": 1.9693082311733804e-05,
"loss": 0.6342,
"step": 1705
},
{
"epoch": 0.29,
"learning_rate": 1.969089316987741e-05,
"loss": 0.4573,
"step": 1710
},
{
"epoch": 0.29,
"learning_rate": 1.9688704028021016e-05,
"loss": 0.2317,
"step": 1715
},
{
"epoch": 0.29,
"learning_rate": 1.9686514886164626e-05,
"loss": 0.2594,
"step": 1720
},
{
"epoch": 0.3,
"learning_rate": 1.9684325744308233e-05,
"loss": 0.292,
"step": 1725
},
{
"epoch": 0.3,
"learning_rate": 1.968213660245184e-05,
"loss": 0.3363,
"step": 1730
},
{
"epoch": 0.3,
"learning_rate": 1.967994746059545e-05,
"loss": 0.344,
"step": 1735
},
{
"epoch": 0.3,
"learning_rate": 1.9677758318739055e-05,
"loss": 0.3528,
"step": 1740
},
{
"epoch": 0.3,
"learning_rate": 1.9675569176882662e-05,
"loss": 0.4914,
"step": 1745
},
{
"epoch": 0.3,
"learning_rate": 1.967338003502627e-05,
"loss": 0.6248,
"step": 1750
},
{
"epoch": 0.3,
"learning_rate": 1.9671190893169878e-05,
"loss": 0.6435,
"step": 1755
},
{
"epoch": 0.3,
"learning_rate": 1.9669001751313485e-05,
"loss": 0.4346,
"step": 1760
},
{
"epoch": 0.3,
"learning_rate": 1.9666812609457094e-05,
"loss": 0.2467,
"step": 1765
},
{
"epoch": 0.3,
"learning_rate": 1.9664623467600704e-05,
"loss": 0.2768,
"step": 1770
},
{
"epoch": 0.3,
"learning_rate": 1.966243432574431e-05,
"loss": 0.2894,
"step": 1775
},
{
"epoch": 0.31,
"learning_rate": 1.9660245183887917e-05,
"loss": 0.3707,
"step": 1780
},
{
"epoch": 0.31,
"learning_rate": 1.9658056042031527e-05,
"loss": 0.389,
"step": 1785
},
{
"epoch": 0.31,
"learning_rate": 1.9655866900175133e-05,
"loss": 0.463,
"step": 1790
},
{
"epoch": 0.31,
"learning_rate": 1.965367775831874e-05,
"loss": 0.5027,
"step": 1795
},
{
"epoch": 0.31,
"learning_rate": 1.965148861646235e-05,
"loss": 0.69,
"step": 1800
},
{
"epoch": 0.31,
"learning_rate": 1.9649299474605956e-05,
"loss": 0.6129,
"step": 1805
},
{
"epoch": 0.31,
"learning_rate": 1.9647110332749562e-05,
"loss": 0.4047,
"step": 1810
},
{
"epoch": 0.31,
"learning_rate": 1.9644921190893172e-05,
"loss": 0.3004,
"step": 1815
},
{
"epoch": 0.31,
"learning_rate": 1.964273204903678e-05,
"loss": 0.2855,
"step": 1820
},
{
"epoch": 0.31,
"learning_rate": 1.9640542907180385e-05,
"loss": 0.3498,
"step": 1825
},
{
"epoch": 0.31,
"learning_rate": 1.9638353765323995e-05,
"loss": 0.288,
"step": 1830
},
{
"epoch": 0.31,
"learning_rate": 1.96361646234676e-05,
"loss": 0.3777,
"step": 1835
},
{
"epoch": 0.32,
"learning_rate": 1.9633975481611208e-05,
"loss": 0.411,
"step": 1840
},
{
"epoch": 0.32,
"learning_rate": 1.9631786339754818e-05,
"loss": 0.4758,
"step": 1845
},
{
"epoch": 0.32,
"learning_rate": 1.9629597197898424e-05,
"loss": 0.6398,
"step": 1850
},
{
"epoch": 0.32,
"learning_rate": 1.962740805604203e-05,
"loss": 0.6373,
"step": 1855
},
{
"epoch": 0.32,
"learning_rate": 1.962521891418564e-05,
"loss": 0.4322,
"step": 1860
},
{
"epoch": 0.32,
"learning_rate": 1.962302977232925e-05,
"loss": 0.2339,
"step": 1865
},
{
"epoch": 0.32,
"learning_rate": 1.9620840630472857e-05,
"loss": 0.2422,
"step": 1870
},
{
"epoch": 0.32,
"learning_rate": 1.9618651488616463e-05,
"loss": 0.327,
"step": 1875
},
{
"epoch": 0.32,
"learning_rate": 1.9616462346760073e-05,
"loss": 0.385,
"step": 1880
},
{
"epoch": 0.32,
"learning_rate": 1.961427320490368e-05,
"loss": 0.3259,
"step": 1885
},
{
"epoch": 0.32,
"learning_rate": 1.961208406304729e-05,
"loss": 0.3767,
"step": 1890
},
{
"epoch": 0.32,
"learning_rate": 1.9609894921190896e-05,
"loss": 0.4718,
"step": 1895
},
{
"epoch": 0.33,
"learning_rate": 1.9607705779334502e-05,
"loss": 0.7808,
"step": 1900
},
{
"epoch": 0.33,
"learning_rate": 1.9605516637478112e-05,
"loss": 0.6041,
"step": 1905
},
{
"epoch": 0.33,
"learning_rate": 1.9603327495621718e-05,
"loss": 0.4006,
"step": 1910
},
{
"epoch": 0.33,
"learning_rate": 1.9601138353765325e-05,
"loss": 0.2423,
"step": 1915
},
{
"epoch": 0.33,
"learning_rate": 1.9598949211908935e-05,
"loss": 0.2883,
"step": 1920
},
{
"epoch": 0.33,
"learning_rate": 1.959676007005254e-05,
"loss": 0.2752,
"step": 1925
},
{
"epoch": 0.33,
"learning_rate": 1.9594570928196147e-05,
"loss": 0.2794,
"step": 1930
},
{
"epoch": 0.33,
"learning_rate": 1.9592381786339757e-05,
"loss": 0.3148,
"step": 1935
},
{
"epoch": 0.33,
"learning_rate": 1.9590192644483364e-05,
"loss": 0.3867,
"step": 1940
},
{
"epoch": 0.33,
"learning_rate": 1.958800350262697e-05,
"loss": 0.5189,
"step": 1945
},
{
"epoch": 0.33,
"learning_rate": 1.958581436077058e-05,
"loss": 0.6908,
"step": 1950
},
{
"epoch": 0.33,
"learning_rate": 1.9583625218914186e-05,
"loss": 0.5684,
"step": 1955
},
{
"epoch": 0.34,
"learning_rate": 1.9581436077057796e-05,
"loss": 0.3987,
"step": 1960
},
{
"epoch": 0.34,
"learning_rate": 1.9579246935201403e-05,
"loss": 0.201,
"step": 1965
},
{
"epoch": 0.34,
"learning_rate": 1.9577057793345012e-05,
"loss": 0.2666,
"step": 1970
},
{
"epoch": 0.34,
"learning_rate": 1.957486865148862e-05,
"loss": 0.2453,
"step": 1975
},
{
"epoch": 0.34,
"learning_rate": 1.9572679509632225e-05,
"loss": 0.4236,
"step": 1980
},
{
"epoch": 0.34,
"learning_rate": 1.9570490367775835e-05,
"loss": 0.3309,
"step": 1985
},
{
"epoch": 0.34,
"learning_rate": 1.956830122591944e-05,
"loss": 0.379,
"step": 1990
},
{
"epoch": 0.34,
"learning_rate": 1.9566112084063048e-05,
"loss": 0.4711,
"step": 1995
},
{
"epoch": 0.34,
"learning_rate": 1.9563922942206658e-05,
"loss": 0.6011,
"step": 2000
},
{
"epoch": 0.34,
"eval_loss": 0.22457240521907806,
"eval_runtime": 1309.2737,
"eval_samples_per_second": 15.848,
"eval_steps_per_second": 3.963,
"eval_wer": 0.7249946455343756,
"step": 2000
},
{
"epoch": 0.34,
"learning_rate": 1.9561733800350264e-05,
"loss": 0.588,
"step": 2005
},
{
"epoch": 0.34,
"learning_rate": 1.955954465849387e-05,
"loss": 0.4,
"step": 2010
},
{
"epoch": 0.35,
"learning_rate": 1.955735551663748e-05,
"loss": 0.1814,
"step": 2015
},
{
"epoch": 0.35,
"learning_rate": 1.9555166374781087e-05,
"loss": 0.2176,
"step": 2020
},
{
"epoch": 0.35,
"learning_rate": 1.9552977232924693e-05,
"loss": 0.2657,
"step": 2025
},
{
"epoch": 0.35,
"learning_rate": 1.9550788091068303e-05,
"loss": 0.3015,
"step": 2030
},
{
"epoch": 0.35,
"learning_rate": 1.954859894921191e-05,
"loss": 0.3371,
"step": 2035
},
{
"epoch": 0.35,
"learning_rate": 1.9546409807355516e-05,
"loss": 0.3518,
"step": 2040
},
{
"epoch": 0.35,
"learning_rate": 1.9544220665499126e-05,
"loss": 0.4323,
"step": 2045
},
{
"epoch": 0.35,
"learning_rate": 1.9542031523642732e-05,
"loss": 0.7322,
"step": 2050
},
{
"epoch": 0.35,
"learning_rate": 1.953984238178634e-05,
"loss": 0.5688,
"step": 2055
},
{
"epoch": 0.35,
"learning_rate": 1.953765323992995e-05,
"loss": 0.4094,
"step": 2060
},
{
"epoch": 0.35,
"learning_rate": 1.953546409807356e-05,
"loss": 0.2227,
"step": 2065
},
{
"epoch": 0.35,
"learning_rate": 1.9533274956217165e-05,
"loss": 0.1978,
"step": 2070
},
{
"epoch": 0.36,
"learning_rate": 1.953108581436077e-05,
"loss": 0.2461,
"step": 2075
},
{
"epoch": 0.36,
"learning_rate": 1.952889667250438e-05,
"loss": 0.316,
"step": 2080
},
{
"epoch": 0.36,
"learning_rate": 1.9526707530647988e-05,
"loss": 0.3647,
"step": 2085
},
{
"epoch": 0.36,
"learning_rate": 1.9524518388791594e-05,
"loss": 0.3506,
"step": 2090
},
{
"epoch": 0.36,
"learning_rate": 1.9522329246935204e-05,
"loss": 0.4282,
"step": 2095
},
{
"epoch": 0.36,
"learning_rate": 1.952014010507881e-05,
"loss": 0.6256,
"step": 2100
},
{
"epoch": 0.36,
"learning_rate": 1.951795096322242e-05,
"loss": 0.6428,
"step": 2105
},
{
"epoch": 0.36,
"learning_rate": 1.9515761821366027e-05,
"loss": 0.3721,
"step": 2110
},
{
"epoch": 0.36,
"learning_rate": 1.9513572679509633e-05,
"loss": 0.227,
"step": 2115
},
{
"epoch": 0.36,
"learning_rate": 1.9511383537653243e-05,
"loss": 0.259,
"step": 2120
},
{
"epoch": 0.36,
"learning_rate": 1.950919439579685e-05,
"loss": 0.2431,
"step": 2125
},
{
"epoch": 0.36,
"learning_rate": 1.9507005253940456e-05,
"loss": 0.3066,
"step": 2130
},
{
"epoch": 0.37,
"learning_rate": 1.9504816112084065e-05,
"loss": 0.3544,
"step": 2135
},
{
"epoch": 0.37,
"learning_rate": 1.9502626970227672e-05,
"loss": 0.3158,
"step": 2140
},
{
"epoch": 0.37,
"learning_rate": 1.950043782837128e-05,
"loss": 0.4208,
"step": 2145
},
{
"epoch": 0.37,
"learning_rate": 1.9498248686514888e-05,
"loss": 0.6774,
"step": 2150
},
{
"epoch": 0.37,
"learning_rate": 1.9496059544658495e-05,
"loss": 0.5467,
"step": 2155
},
{
"epoch": 0.37,
"learning_rate": 1.9493870402802104e-05,
"loss": 0.4307,
"step": 2160
},
{
"epoch": 0.37,
"learning_rate": 1.949168126094571e-05,
"loss": 0.2036,
"step": 2165
},
{
"epoch": 0.37,
"learning_rate": 1.948949211908932e-05,
"loss": 0.2138,
"step": 2170
},
{
"epoch": 0.37,
"learning_rate": 1.9487302977232927e-05,
"loss": 0.2442,
"step": 2175
},
{
"epoch": 0.37,
"learning_rate": 1.9485113835376534e-05,
"loss": 0.3362,
"step": 2180
},
{
"epoch": 0.37,
"learning_rate": 1.9482924693520143e-05,
"loss": 0.4253,
"step": 2185
},
{
"epoch": 0.38,
"learning_rate": 1.948073555166375e-05,
"loss": 0.3091,
"step": 2190
},
{
"epoch": 0.38,
"learning_rate": 1.9478546409807356e-05,
"loss": 0.399,
"step": 2195
},
{
"epoch": 0.38,
"learning_rate": 1.9476357267950966e-05,
"loss": 0.6479,
"step": 2200
},
{
"epoch": 0.38,
"learning_rate": 1.9474168126094573e-05,
"loss": 0.5511,
"step": 2205
},
{
"epoch": 0.38,
"learning_rate": 1.947197898423818e-05,
"loss": 0.465,
"step": 2210
},
{
"epoch": 0.38,
"learning_rate": 1.946978984238179e-05,
"loss": 0.1555,
"step": 2215
},
{
"epoch": 0.38,
"learning_rate": 1.9467600700525395e-05,
"loss": 0.1958,
"step": 2220
},
{
"epoch": 0.38,
"learning_rate": 1.9465411558669e-05,
"loss": 0.2306,
"step": 2225
},
{
"epoch": 0.38,
"learning_rate": 1.946322241681261e-05,
"loss": 0.2869,
"step": 2230
},
{
"epoch": 0.38,
"learning_rate": 1.9461033274956218e-05,
"loss": 0.2702,
"step": 2235
},
{
"epoch": 0.38,
"learning_rate": 1.9458844133099824e-05,
"loss": 0.3525,
"step": 2240
},
{
"epoch": 0.38,
"learning_rate": 1.9456654991243434e-05,
"loss": 0.4003,
"step": 2245
},
{
"epoch": 0.39,
"learning_rate": 1.945446584938704e-05,
"loss": 0.6201,
"step": 2250
},
{
"epoch": 0.39,
"learning_rate": 1.945227670753065e-05,
"loss": 0.6106,
"step": 2255
},
{
"epoch": 0.39,
"learning_rate": 1.9450087565674257e-05,
"loss": 0.3657,
"step": 2260
},
{
"epoch": 0.39,
"learning_rate": 1.9447898423817867e-05,
"loss": 0.2049,
"step": 2265
},
{
"epoch": 0.39,
"learning_rate": 1.9445709281961473e-05,
"loss": 0.2005,
"step": 2270
},
{
"epoch": 0.39,
"learning_rate": 1.944352014010508e-05,
"loss": 0.2183,
"step": 2275
},
{
"epoch": 0.39,
"learning_rate": 1.944133099824869e-05,
"loss": 0.3148,
"step": 2280
},
{
"epoch": 0.39,
"learning_rate": 1.9439141856392296e-05,
"loss": 0.3327,
"step": 2285
},
{
"epoch": 0.39,
"learning_rate": 1.9436952714535902e-05,
"loss": 0.4217,
"step": 2290
},
{
"epoch": 0.39,
"learning_rate": 1.9434763572679512e-05,
"loss": 0.3703,
"step": 2295
},
{
"epoch": 0.39,
"learning_rate": 1.943257443082312e-05,
"loss": 0.6257,
"step": 2300
},
{
"epoch": 0.39,
"learning_rate": 1.9430385288966725e-05,
"loss": 0.5866,
"step": 2305
},
{
"epoch": 0.4,
"learning_rate": 1.9428196147110335e-05,
"loss": 0.4075,
"step": 2310
},
{
"epoch": 0.4,
"learning_rate": 1.942600700525394e-05,
"loss": 0.2202,
"step": 2315
},
{
"epoch": 0.4,
"learning_rate": 1.942381786339755e-05,
"loss": 0.1691,
"step": 2320
},
{
"epoch": 0.4,
"learning_rate": 1.9421628721541157e-05,
"loss": 0.2135,
"step": 2325
},
{
"epoch": 0.4,
"learning_rate": 1.9419439579684764e-05,
"loss": 0.2311,
"step": 2330
},
{
"epoch": 0.4,
"learning_rate": 1.9417250437828374e-05,
"loss": 0.3002,
"step": 2335
},
{
"epoch": 0.4,
"learning_rate": 1.941506129597198e-05,
"loss": 0.308,
"step": 2340
},
{
"epoch": 0.4,
"learning_rate": 1.9412872154115587e-05,
"loss": 0.4534,
"step": 2345
},
{
"epoch": 0.4,
"learning_rate": 1.9410683012259196e-05,
"loss": 0.5574,
"step": 2350
},
{
"epoch": 0.4,
"learning_rate": 1.9408493870402806e-05,
"loss": 0.5178,
"step": 2355
},
{
"epoch": 0.4,
"learning_rate": 1.9406304728546413e-05,
"loss": 0.3753,
"step": 2360
},
{
"epoch": 0.41,
"learning_rate": 1.940411558669002e-05,
"loss": 0.1951,
"step": 2365
},
{
"epoch": 0.41,
"learning_rate": 1.940192644483363e-05,
"loss": 0.1972,
"step": 2370
},
{
"epoch": 0.41,
"learning_rate": 1.9399737302977235e-05,
"loss": 0.2758,
"step": 2375
},
{
"epoch": 0.41,
"learning_rate": 1.9397548161120842e-05,
"loss": 0.2595,
"step": 2380
},
{
"epoch": 0.41,
"learning_rate": 1.939535901926445e-05,
"loss": 0.321,
"step": 2385
},
{
"epoch": 0.41,
"learning_rate": 1.9393169877408058e-05,
"loss": 0.2839,
"step": 2390
},
{
"epoch": 0.41,
"learning_rate": 1.9390980735551665e-05,
"loss": 0.418,
"step": 2395
},
{
"epoch": 0.41,
"learning_rate": 1.9388791593695274e-05,
"loss": 0.6943,
"step": 2400
},
{
"epoch": 0.41,
"learning_rate": 1.938660245183888e-05,
"loss": 0.526,
"step": 2405
},
{
"epoch": 0.41,
"learning_rate": 1.9384413309982487e-05,
"loss": 0.3193,
"step": 2410
},
{
"epoch": 0.41,
"learning_rate": 1.9382224168126097e-05,
"loss": 0.2126,
"step": 2415
},
{
"epoch": 0.41,
"learning_rate": 1.9380035026269703e-05,
"loss": 0.2117,
"step": 2420
},
{
"epoch": 0.42,
"learning_rate": 1.937784588441331e-05,
"loss": 0.2111,
"step": 2425
},
{
"epoch": 0.42,
"learning_rate": 1.937565674255692e-05,
"loss": 0.3039,
"step": 2430
},
{
"epoch": 0.42,
"learning_rate": 1.9373467600700526e-05,
"loss": 0.3009,
"step": 2435
},
{
"epoch": 0.42,
"learning_rate": 1.9371278458844133e-05,
"loss": 0.3634,
"step": 2440
},
{
"epoch": 0.42,
"learning_rate": 1.9369089316987742e-05,
"loss": 0.4278,
"step": 2445
},
{
"epoch": 0.42,
"learning_rate": 1.936690017513135e-05,
"loss": 0.6185,
"step": 2450
},
{
"epoch": 0.42,
"learning_rate": 1.936471103327496e-05,
"loss": 0.5109,
"step": 2455
},
{
"epoch": 0.42,
"learning_rate": 1.9362521891418565e-05,
"loss": 0.3568,
"step": 2460
},
{
"epoch": 0.42,
"learning_rate": 1.9360332749562175e-05,
"loss": 0.2025,
"step": 2465
},
{
"epoch": 0.42,
"learning_rate": 1.935814360770578e-05,
"loss": 0.2216,
"step": 2470
},
{
"epoch": 0.42,
"learning_rate": 1.9355954465849388e-05,
"loss": 0.2445,
"step": 2475
},
{
"epoch": 0.42,
"learning_rate": 1.9353765323992998e-05,
"loss": 0.281,
"step": 2480
},
{
"epoch": 0.43,
"learning_rate": 1.9351576182136604e-05,
"loss": 0.2977,
"step": 2485
},
{
"epoch": 0.43,
"learning_rate": 1.934938704028021e-05,
"loss": 0.3259,
"step": 2490
},
{
"epoch": 0.43,
"learning_rate": 1.934719789842382e-05,
"loss": 0.3275,
"step": 2495
},
{
"epoch": 0.43,
"learning_rate": 1.9345008756567427e-05,
"loss": 0.6187,
"step": 2500
},
{
"epoch": 0.43,
"eval_loss": 0.11373376101255417,
"eval_runtime": 1311.5923,
"eval_samples_per_second": 15.82,
"eval_steps_per_second": 3.955,
"eval_wer": 0.5098522167487685,
"step": 2500
},
{
"epoch": 0.43,
"learning_rate": 1.9342819614711033e-05,
"loss": 0.5066,
"step": 2505
},
{
"epoch": 0.43,
"learning_rate": 1.9340630472854643e-05,
"loss": 0.3058,
"step": 2510
},
{
"epoch": 0.43,
"learning_rate": 1.933844133099825e-05,
"loss": 0.1477,
"step": 2515
},
{
"epoch": 0.43,
"learning_rate": 1.9336252189141856e-05,
"loss": 0.2596,
"step": 2520
},
{
"epoch": 0.43,
"learning_rate": 1.9334063047285466e-05,
"loss": 0.2404,
"step": 2525
},
{
"epoch": 0.43,
"learning_rate": 1.9331873905429072e-05,
"loss": 0.2375,
"step": 2530
},
{
"epoch": 0.43,
"learning_rate": 1.932968476357268e-05,
"loss": 0.3477,
"step": 2535
},
{
"epoch": 0.44,
"learning_rate": 1.932749562171629e-05,
"loss": 0.3373,
"step": 2540
},
{
"epoch": 0.44,
"learning_rate": 1.9325306479859895e-05,
"loss": 0.3746,
"step": 2545
},
{
"epoch": 0.44,
"learning_rate": 1.9323117338003505e-05,
"loss": 0.5801,
"step": 2550
},
{
"epoch": 0.44,
"learning_rate": 1.9320928196147115e-05,
"loss": 0.5143,
"step": 2555
},
{
"epoch": 0.44,
"learning_rate": 1.931873905429072e-05,
"loss": 0.3353,
"step": 2560
},
{
"epoch": 0.44,
"learning_rate": 1.9316549912434327e-05,
"loss": 0.1744,
"step": 2565
},
{
"epoch": 0.44,
"learning_rate": 1.9314360770577937e-05,
"loss": 0.2012,
"step": 2570
},
{
"epoch": 0.44,
"learning_rate": 1.9312171628721544e-05,
"loss": 0.2695,
"step": 2575
},
{
"epoch": 0.44,
"learning_rate": 1.930998248686515e-05,
"loss": 0.3254,
"step": 2580
},
{
"epoch": 0.44,
"learning_rate": 1.930779334500876e-05,
"loss": 0.2918,
"step": 2585
},
{
"epoch": 0.44,
"learning_rate": 1.9305604203152366e-05,
"loss": 0.3399,
"step": 2590
},
{
"epoch": 0.44,
"learning_rate": 1.9303415061295973e-05,
"loss": 0.3886,
"step": 2595
},
{
"epoch": 0.45,
"learning_rate": 1.9301225919439583e-05,
"loss": 0.5188,
"step": 2600
},
{
"epoch": 0.45,
"learning_rate": 1.929903677758319e-05,
"loss": 0.5015,
"step": 2605
},
{
"epoch": 0.45,
"learning_rate": 1.9296847635726796e-05,
"loss": 0.3697,
"step": 2610
},
{
"epoch": 0.45,
"learning_rate": 1.9294658493870405e-05,
"loss": 0.1413,
"step": 2615
},
{
"epoch": 0.45,
"learning_rate": 1.9292469352014012e-05,
"loss": 0.2042,
"step": 2620
},
{
"epoch": 0.45,
"learning_rate": 1.9290280210157618e-05,
"loss": 0.235,
"step": 2625
},
{
"epoch": 0.45,
"learning_rate": 1.9288091068301228e-05,
"loss": 0.2638,
"step": 2630
},
{
"epoch": 0.45,
"learning_rate": 1.9285901926444834e-05,
"loss": 0.3261,
"step": 2635
},
{
"epoch": 0.45,
"learning_rate": 1.928371278458844e-05,
"loss": 0.2925,
"step": 2640
},
{
"epoch": 0.45,
"learning_rate": 1.928152364273205e-05,
"loss": 0.3494,
"step": 2645
},
{
"epoch": 0.45,
"learning_rate": 1.927933450087566e-05,
"loss": 0.6889,
"step": 2650
},
{
"epoch": 0.45,
"learning_rate": 1.9277145359019267e-05,
"loss": 0.478,
"step": 2655
},
{
"epoch": 0.46,
"learning_rate": 1.9274956217162873e-05,
"loss": 0.2939,
"step": 2660
},
{
"epoch": 0.46,
"learning_rate": 1.9272767075306483e-05,
"loss": 0.1574,
"step": 2665
},
{
"epoch": 0.46,
"learning_rate": 1.927057793345009e-05,
"loss": 0.2173,
"step": 2670
},
{
"epoch": 0.46,
"learning_rate": 1.9268388791593696e-05,
"loss": 0.2178,
"step": 2675
},
{
"epoch": 0.46,
"learning_rate": 1.9266199649737306e-05,
"loss": 0.2264,
"step": 2680
},
{
"epoch": 0.46,
"learning_rate": 1.9264010507880912e-05,
"loss": 0.2803,
"step": 2685
},
{
"epoch": 0.46,
"learning_rate": 1.926182136602452e-05,
"loss": 0.3286,
"step": 2690
},
{
"epoch": 0.46,
"learning_rate": 1.925963222416813e-05,
"loss": 0.3679,
"step": 2695
},
{
"epoch": 0.46,
"learning_rate": 1.9257443082311735e-05,
"loss": 0.5848,
"step": 2700
},
{
"epoch": 0.46,
"learning_rate": 1.925525394045534e-05,
"loss": 0.5183,
"step": 2705
},
{
"epoch": 0.46,
"learning_rate": 1.925306479859895e-05,
"loss": 0.2741,
"step": 2710
},
{
"epoch": 0.47,
"learning_rate": 1.9250875656742558e-05,
"loss": 0.1843,
"step": 2715
},
{
"epoch": 0.47,
"learning_rate": 1.9248686514886164e-05,
"loss": 0.2052,
"step": 2720
},
{
"epoch": 0.47,
"learning_rate": 1.9246497373029774e-05,
"loss": 0.2062,
"step": 2725
},
{
"epoch": 0.47,
"learning_rate": 1.924430823117338e-05,
"loss": 0.2863,
"step": 2730
},
{
"epoch": 0.47,
"learning_rate": 1.9242119089316987e-05,
"loss": 0.2867,
"step": 2735
},
{
"epoch": 0.47,
"learning_rate": 1.9239929947460597e-05,
"loss": 0.3428,
"step": 2740
},
{
"epoch": 0.47,
"learning_rate": 1.9237740805604203e-05,
"loss": 0.399,
"step": 2745
},
{
"epoch": 0.47,
"learning_rate": 1.9235551663747813e-05,
"loss": 0.614,
"step": 2750
},
{
"epoch": 0.47,
"learning_rate": 1.923336252189142e-05,
"loss": 0.4946,
"step": 2755
},
{
"epoch": 0.47,
"learning_rate": 1.923117338003503e-05,
"loss": 0.2238,
"step": 2760
},
{
"epoch": 0.47,
"learning_rate": 1.9228984238178636e-05,
"loss": 0.1538,
"step": 2765
},
{
"epoch": 0.47,
"learning_rate": 1.9226795096322246e-05,
"loss": 0.1702,
"step": 2770
},
{
"epoch": 0.48,
"learning_rate": 1.9224605954465852e-05,
"loss": 0.2204,
"step": 2775
},
{
"epoch": 0.48,
"learning_rate": 1.922241681260946e-05,
"loss": 0.2294,
"step": 2780
},
{
"epoch": 0.48,
"learning_rate": 1.9220227670753068e-05,
"loss": 0.3237,
"step": 2785
},
{
"epoch": 0.48,
"learning_rate": 1.9218038528896675e-05,
"loss": 0.3188,
"step": 2790
},
{
"epoch": 0.48,
"learning_rate": 1.921584938704028e-05,
"loss": 0.3757,
"step": 2795
},
{
"epoch": 0.48,
"learning_rate": 1.921366024518389e-05,
"loss": 0.5675,
"step": 2800
},
{
"epoch": 0.48,
"learning_rate": 1.9211471103327497e-05,
"loss": 0.5227,
"step": 2805
},
{
"epoch": 0.48,
"learning_rate": 1.9209281961471104e-05,
"loss": 0.3396,
"step": 2810
},
{
"epoch": 0.48,
"learning_rate": 1.9207092819614714e-05,
"loss": 0.1774,
"step": 2815
},
{
"epoch": 0.48,
"learning_rate": 1.920490367775832e-05,
"loss": 0.1719,
"step": 2820
},
{
"epoch": 0.48,
"learning_rate": 1.9202714535901926e-05,
"loss": 0.1772,
"step": 2825
},
{
"epoch": 0.48,
"learning_rate": 1.9200525394045536e-05,
"loss": 0.2546,
"step": 2830
},
{
"epoch": 0.49,
"learning_rate": 1.9198336252189143e-05,
"loss": 0.2397,
"step": 2835
},
{
"epoch": 0.49,
"learning_rate": 1.919614711033275e-05,
"loss": 0.316,
"step": 2840
},
{
"epoch": 0.49,
"learning_rate": 1.919395796847636e-05,
"loss": 0.4364,
"step": 2845
},
{
"epoch": 0.49,
"learning_rate": 1.919176882661997e-05,
"loss": 0.4873,
"step": 2850
},
{
"epoch": 0.49,
"learning_rate": 1.9189579684763575e-05,
"loss": 0.4671,
"step": 2855
},
{
"epoch": 0.49,
"learning_rate": 1.918739054290718e-05,
"loss": 0.3178,
"step": 2860
},
{
"epoch": 0.49,
"learning_rate": 1.918520140105079e-05,
"loss": 0.1546,
"step": 2865
},
{
"epoch": 0.49,
"learning_rate": 1.9183012259194398e-05,
"loss": 0.1624,
"step": 2870
},
{
"epoch": 0.49,
"learning_rate": 1.9180823117338004e-05,
"loss": 0.181,
"step": 2875
},
{
"epoch": 0.49,
"learning_rate": 1.9178633975481614e-05,
"loss": 0.22,
"step": 2880
},
{
"epoch": 0.49,
"learning_rate": 1.917644483362522e-05,
"loss": 0.2576,
"step": 2885
},
{
"epoch": 0.5,
"learning_rate": 1.9174255691768827e-05,
"loss": 0.3094,
"step": 2890
},
{
"epoch": 0.5,
"learning_rate": 1.9172066549912437e-05,
"loss": 0.35,
"step": 2895
},
{
"epoch": 0.5,
"learning_rate": 1.9169877408056043e-05,
"loss": 0.5505,
"step": 2900
},
{
"epoch": 0.5,
"learning_rate": 1.916768826619965e-05,
"loss": 0.4483,
"step": 2905
},
{
"epoch": 0.5,
"learning_rate": 1.916549912434326e-05,
"loss": 0.2565,
"step": 2910
},
{
"epoch": 0.5,
"learning_rate": 1.9163309982486866e-05,
"loss": 0.1526,
"step": 2915
},
{
"epoch": 0.5,
"learning_rate": 1.9161120840630472e-05,
"loss": 0.1785,
"step": 2920
},
{
"epoch": 0.5,
"learning_rate": 1.9158931698774082e-05,
"loss": 0.2053,
"step": 2925
},
{
"epoch": 0.5,
"learning_rate": 1.915674255691769e-05,
"loss": 0.2105,
"step": 2930
},
{
"epoch": 0.5,
"learning_rate": 1.9154553415061295e-05,
"loss": 0.2896,
"step": 2935
},
{
"epoch": 0.5,
"learning_rate": 1.9152364273204905e-05,
"loss": 0.3235,
"step": 2940
},
{
"epoch": 0.5,
"learning_rate": 1.9150175131348515e-05,
"loss": 0.4154,
"step": 2945
},
{
"epoch": 0.51,
"learning_rate": 1.914798598949212e-05,
"loss": 0.4711,
"step": 2950
},
{
"epoch": 0.51,
"learning_rate": 1.9145796847635728e-05,
"loss": 0.482,
"step": 2955
},
{
"epoch": 0.51,
"learning_rate": 1.9143607705779338e-05,
"loss": 0.359,
"step": 2960
},
{
"epoch": 0.51,
"learning_rate": 1.9141418563922944e-05,
"loss": 0.1506,
"step": 2965
},
{
"epoch": 0.51,
"learning_rate": 1.913922942206655e-05,
"loss": 0.1876,
"step": 2970
},
{
"epoch": 0.51,
"learning_rate": 1.913704028021016e-05,
"loss": 0.2387,
"step": 2975
},
{
"epoch": 0.51,
"learning_rate": 1.9134851138353767e-05,
"loss": 0.2472,
"step": 2980
},
{
"epoch": 0.51,
"learning_rate": 1.9132661996497373e-05,
"loss": 0.2706,
"step": 2985
},
{
"epoch": 0.51,
"learning_rate": 1.9130472854640983e-05,
"loss": 0.2585,
"step": 2990
},
{
"epoch": 0.51,
"learning_rate": 1.912828371278459e-05,
"loss": 0.3663,
"step": 2995
},
{
"epoch": 0.51,
"learning_rate": 1.91260945709282e-05,
"loss": 0.5902,
"step": 3000
},
{
"epoch": 0.51,
"eval_loss": 0.12195305526256561,
"eval_runtime": 1676.684,
"eval_samples_per_second": 12.376,
"eval_steps_per_second": 3.094,
"eval_wer": 0.43759370314842577,
"step": 3000
},
{
"epoch": 0.51,
"learning_rate": 1.9123905429071806e-05,
"loss": 0.4624,
"step": 3005
},
{
"epoch": 0.52,
"learning_rate": 1.9121716287215412e-05,
"loss": 0.3451,
"step": 3010
},
{
"epoch": 0.52,
"learning_rate": 1.9119527145359022e-05,
"loss": 0.1515,
"step": 3015
},
{
"epoch": 0.52,
"learning_rate": 1.9117338003502628e-05,
"loss": 0.1756,
"step": 3020
},
{
"epoch": 0.52,
"learning_rate": 1.9115148861646235e-05,
"loss": 0.1999,
"step": 3025
},
{
"epoch": 0.52,
"learning_rate": 1.9112959719789845e-05,
"loss": 0.2947,
"step": 3030
},
{
"epoch": 0.52,
"learning_rate": 1.911077057793345e-05,
"loss": 0.2687,
"step": 3035
},
{
"epoch": 0.52,
"learning_rate": 1.9108581436077057e-05,
"loss": 0.2746,
"step": 3040
},
{
"epoch": 0.52,
"learning_rate": 1.9106392294220667e-05,
"loss": 0.3335,
"step": 3045
},
{
"epoch": 0.52,
"learning_rate": 1.9104203152364277e-05,
"loss": 0.5772,
"step": 3050
},
{
"epoch": 0.52,
"learning_rate": 1.9102014010507884e-05,
"loss": 0.4858,
"step": 3055
},
{
"epoch": 0.52,
"learning_rate": 1.909982486865149e-05,
"loss": 0.3047,
"step": 3060
},
{
"epoch": 0.53,
"learning_rate": 1.90976357267951e-05,
"loss": 0.1478,
"step": 3065
},
{
"epoch": 0.53,
"learning_rate": 1.9095446584938706e-05,
"loss": 0.196,
"step": 3070
},
{
"epoch": 0.53,
"learning_rate": 1.9093257443082313e-05,
"loss": 0.1793,
"step": 3075
},
{
"epoch": 0.53,
"learning_rate": 1.9091068301225922e-05,
"loss": 0.2277,
"step": 3080
},
{
"epoch": 0.53,
"learning_rate": 1.908887915936953e-05,
"loss": 0.2503,
"step": 3085
},
{
"epoch": 0.53,
"learning_rate": 1.9086690017513135e-05,
"loss": 0.3468,
"step": 3090
},
{
"epoch": 0.53,
"learning_rate": 1.9084500875656745e-05,
"loss": 0.3171,
"step": 3095
},
{
"epoch": 0.53,
"learning_rate": 1.908231173380035e-05,
"loss": 0.5396,
"step": 3100
},
{
"epoch": 0.53,
"learning_rate": 1.9080122591943958e-05,
"loss": 0.4492,
"step": 3105
},
{
"epoch": 0.53,
"learning_rate": 1.9077933450087568e-05,
"loss": 0.254,
"step": 3110
},
{
"epoch": 0.53,
"learning_rate": 1.9075744308231174e-05,
"loss": 0.1517,
"step": 3115
},
{
"epoch": 0.53,
"learning_rate": 1.907355516637478e-05,
"loss": 0.1726,
"step": 3120
},
{
"epoch": 0.54,
"learning_rate": 1.907136602451839e-05,
"loss": 0.2011,
"step": 3125
},
{
"epoch": 0.54,
"learning_rate": 1.9069176882661997e-05,
"loss": 0.1957,
"step": 3130
},
{
"epoch": 0.54,
"learning_rate": 1.9066987740805603e-05,
"loss": 0.2987,
"step": 3135
},
{
"epoch": 0.54,
"learning_rate": 1.9064798598949213e-05,
"loss": 0.3125,
"step": 3140
},
{
"epoch": 0.54,
"learning_rate": 1.9062609457092823e-05,
"loss": 0.3696,
"step": 3145
},
{
"epoch": 0.54,
"learning_rate": 1.906042031523643e-05,
"loss": 0.5477,
"step": 3150
},
{
"epoch": 0.54,
"learning_rate": 1.9058231173380036e-05,
"loss": 0.4396,
"step": 3155
},
{
"epoch": 0.54,
"learning_rate": 1.9056042031523646e-05,
"loss": 0.3259,
"step": 3160
},
{
"epoch": 0.54,
"learning_rate": 1.9053852889667252e-05,
"loss": 0.1482,
"step": 3165
},
{
"epoch": 0.54,
"learning_rate": 1.905166374781086e-05,
"loss": 0.1941,
"step": 3170
},
{
"epoch": 0.54,
"learning_rate": 1.904947460595447e-05,
"loss": 0.1942,
"step": 3175
},
{
"epoch": 0.54,
"learning_rate": 1.9047285464098075e-05,
"loss": 0.2314,
"step": 3180
},
{
"epoch": 0.55,
"learning_rate": 1.904509632224168e-05,
"loss": 0.2693,
"step": 3185
},
{
"epoch": 0.55,
"learning_rate": 1.904290718038529e-05,
"loss": 0.2822,
"step": 3190
},
{
"epoch": 0.55,
"learning_rate": 1.9040718038528898e-05,
"loss": 0.3766,
"step": 3195
},
{
"epoch": 0.55,
"learning_rate": 1.9038528896672504e-05,
"loss": 0.5654,
"step": 3200
},
{
"epoch": 0.55,
"learning_rate": 1.9036339754816114e-05,
"loss": 0.425,
"step": 3205
},
{
"epoch": 0.55,
"learning_rate": 1.903415061295972e-05,
"loss": 0.3129,
"step": 3210
},
{
"epoch": 0.55,
"learning_rate": 1.903196147110333e-05,
"loss": 0.1139,
"step": 3215
},
{
"epoch": 0.55,
"learning_rate": 1.9029772329246937e-05,
"loss": 0.1722,
"step": 3220
},
{
"epoch": 0.55,
"learning_rate": 1.9027583187390543e-05,
"loss": 0.195,
"step": 3225
},
{
"epoch": 0.55,
"learning_rate": 1.9025394045534153e-05,
"loss": 0.215,
"step": 3230
},
{
"epoch": 0.55,
"learning_rate": 1.902320490367776e-05,
"loss": 0.2789,
"step": 3235
},
{
"epoch": 0.56,
"learning_rate": 1.902101576182137e-05,
"loss": 0.2688,
"step": 3240
},
{
"epoch": 0.56,
"learning_rate": 1.9018826619964976e-05,
"loss": 0.3391,
"step": 3245
},
{
"epoch": 0.56,
"learning_rate": 1.9016637478108585e-05,
"loss": 0.5127,
"step": 3250
},
{
"epoch": 0.56,
"learning_rate": 1.9014448336252192e-05,
"loss": 0.4394,
"step": 3255
},
{
"epoch": 0.56,
"learning_rate": 1.9012259194395798e-05,
"loss": 0.3074,
"step": 3260
},
{
"epoch": 0.56,
"learning_rate": 1.9010070052539408e-05,
"loss": 0.1239,
"step": 3265
},
{
"epoch": 0.56,
"learning_rate": 1.9007880910683014e-05,
"loss": 0.1951,
"step": 3270
},
{
"epoch": 0.56,
"learning_rate": 1.900569176882662e-05,
"loss": 0.172,
"step": 3275
},
{
"epoch": 0.56,
"learning_rate": 1.900350262697023e-05,
"loss": 0.2218,
"step": 3280
},
{
"epoch": 0.56,
"learning_rate": 1.9001313485113837e-05,
"loss": 0.2474,
"step": 3285
},
{
"epoch": 0.56,
"learning_rate": 1.8999124343257444e-05,
"loss": 0.2719,
"step": 3290
},
{
"epoch": 0.56,
"learning_rate": 1.8996935201401053e-05,
"loss": 0.3767,
"step": 3295
},
{
"epoch": 0.57,
"learning_rate": 1.899474605954466e-05,
"loss": 0.4909,
"step": 3300
},
{
"epoch": 0.57,
"learning_rate": 1.8992556917688266e-05,
"loss": 0.486,
"step": 3305
},
{
"epoch": 0.57,
"learning_rate": 1.8990367775831876e-05,
"loss": 0.3544,
"step": 3310
},
{
"epoch": 0.57,
"learning_rate": 1.8988178633975483e-05,
"loss": 0.1669,
"step": 3315
},
{
"epoch": 0.57,
"learning_rate": 1.898598949211909e-05,
"loss": 0.1893,
"step": 3320
},
{
"epoch": 0.57,
"learning_rate": 1.89838003502627e-05,
"loss": 0.1736,
"step": 3325
},
{
"epoch": 0.57,
"learning_rate": 1.8981611208406305e-05,
"loss": 0.2142,
"step": 3330
},
{
"epoch": 0.57,
"learning_rate": 1.8979422066549912e-05,
"loss": 0.2013,
"step": 3335
},
{
"epoch": 0.57,
"learning_rate": 1.897723292469352e-05,
"loss": 0.2393,
"step": 3340
},
{
"epoch": 0.57,
"learning_rate": 1.897504378283713e-05,
"loss": 0.3198,
"step": 3345
},
{
"epoch": 0.57,
"learning_rate": 1.8972854640980738e-05,
"loss": 0.5464,
"step": 3350
},
{
"epoch": 0.57,
"learning_rate": 1.8970665499124344e-05,
"loss": 0.398,
"step": 3355
},
{
"epoch": 0.58,
"learning_rate": 1.8968476357267954e-05,
"loss": 0.3023,
"step": 3360
},
{
"epoch": 0.58,
"learning_rate": 1.896628721541156e-05,
"loss": 0.1393,
"step": 3365
},
{
"epoch": 0.58,
"learning_rate": 1.8964098073555167e-05,
"loss": 0.1655,
"step": 3370
},
{
"epoch": 0.58,
"learning_rate": 1.8961908931698777e-05,
"loss": 0.1887,
"step": 3375
},
{
"epoch": 0.58,
"learning_rate": 1.8959719789842383e-05,
"loss": 0.2439,
"step": 3380
},
{
"epoch": 0.58,
"learning_rate": 1.895753064798599e-05,
"loss": 0.2401,
"step": 3385
},
{
"epoch": 0.58,
"learning_rate": 1.89553415061296e-05,
"loss": 0.2706,
"step": 3390
},
{
"epoch": 0.58,
"learning_rate": 1.8953152364273206e-05,
"loss": 0.2931,
"step": 3395
},
{
"epoch": 0.58,
"learning_rate": 1.8950963222416812e-05,
"loss": 0.5721,
"step": 3400
},
{
"epoch": 0.58,
"learning_rate": 1.8948774080560422e-05,
"loss": 0.4469,
"step": 3405
},
{
"epoch": 0.58,
"learning_rate": 1.894658493870403e-05,
"loss": 0.3009,
"step": 3410
},
{
"epoch": 0.59,
"learning_rate": 1.8944395796847635e-05,
"loss": 0.1359,
"step": 3415
},
{
"epoch": 0.59,
"learning_rate": 1.8942206654991245e-05,
"loss": 0.1654,
"step": 3420
},
{
"epoch": 0.59,
"learning_rate": 1.894001751313485e-05,
"loss": 0.2193,
"step": 3425
},
{
"epoch": 0.59,
"learning_rate": 1.893782837127846e-05,
"loss": 0.1865,
"step": 3430
},
{
"epoch": 0.59,
"learning_rate": 1.8935639229422068e-05,
"loss": 0.2639,
"step": 3435
},
{
"epoch": 0.59,
"learning_rate": 1.8933450087565677e-05,
"loss": 0.2666,
"step": 3440
},
{
"epoch": 0.59,
"learning_rate": 1.8931260945709284e-05,
"loss": 0.3394,
"step": 3445
},
{
"epoch": 0.59,
"learning_rate": 1.8929071803852894e-05,
"loss": 0.5245,
"step": 3450
},
{
"epoch": 0.59,
"learning_rate": 1.89268826619965e-05,
"loss": 0.4308,
"step": 3455
},
{
"epoch": 0.59,
"learning_rate": 1.8924693520140106e-05,
"loss": 0.3093,
"step": 3460
},
{
"epoch": 0.59,
"learning_rate": 1.8922504378283716e-05,
"loss": 0.1233,
"step": 3465
},
{
"epoch": 0.59,
"learning_rate": 1.8920315236427323e-05,
"loss": 0.1234,
"step": 3470
},
{
"epoch": 0.6,
"learning_rate": 1.891812609457093e-05,
"loss": 0.1582,
"step": 3475
},
{
"epoch": 0.6,
"learning_rate": 1.891593695271454e-05,
"loss": 0.1989,
"step": 3480
},
{
"epoch": 0.6,
"learning_rate": 1.8913747810858145e-05,
"loss": 0.1945,
"step": 3485
},
{
"epoch": 0.6,
"learning_rate": 1.8911558669001752e-05,
"loss": 0.2911,
"step": 3490
},
{
"epoch": 0.6,
"learning_rate": 1.8909369527145362e-05,
"loss": 0.2994,
"step": 3495
},
{
"epoch": 0.6,
"learning_rate": 1.8907180385288968e-05,
"loss": 0.51,
"step": 3500
},
{
"epoch": 0.6,
"eval_loss": 0.13179369270801544,
"eval_runtime": 1599.8676,
"eval_samples_per_second": 12.97,
"eval_steps_per_second": 3.243,
"eval_wer": 0.41858535018205184,
"step": 3500
},
{
"epoch": 0.6,
"learning_rate": 1.8904991243432575e-05,
"loss": 0.4476,
"step": 3505
},
{
"epoch": 0.6,
"learning_rate": 1.8902802101576184e-05,
"loss": 0.2828,
"step": 3510
},
{
"epoch": 0.6,
"learning_rate": 1.890061295971979e-05,
"loss": 0.0974,
"step": 3515
},
{
"epoch": 0.6,
"learning_rate": 1.8898423817863397e-05,
"loss": 0.1477,
"step": 3520
},
{
"epoch": 0.6,
"learning_rate": 1.8896234676007007e-05,
"loss": 0.1831,
"step": 3525
},
{
"epoch": 0.6,
"learning_rate": 1.8894045534150614e-05,
"loss": 0.2379,
"step": 3530
},
{
"epoch": 0.61,
"learning_rate": 1.8891856392294223e-05,
"loss": 0.196,
"step": 3535
},
{
"epoch": 0.61,
"learning_rate": 1.888966725043783e-05,
"loss": 0.2322,
"step": 3540
},
{
"epoch": 0.61,
"learning_rate": 1.888747810858144e-05,
"loss": 0.3072,
"step": 3545
},
{
"epoch": 0.61,
"learning_rate": 1.8885288966725046e-05,
"loss": 0.4667,
"step": 3550
},
{
"epoch": 0.61,
"learning_rate": 1.8883099824868652e-05,
"loss": 0.4226,
"step": 3555
},
{
"epoch": 0.61,
"learning_rate": 1.8880910683012262e-05,
"loss": 0.2845,
"step": 3560
},
{
"epoch": 0.61,
"learning_rate": 1.887872154115587e-05,
"loss": 0.1823,
"step": 3565
},
{
"epoch": 0.61,
"learning_rate": 1.8876532399299475e-05,
"loss": 0.1303,
"step": 3570
},
{
"epoch": 0.61,
"learning_rate": 1.8874343257443085e-05,
"loss": 0.1759,
"step": 3575
},
{
"epoch": 0.61,
"learning_rate": 1.887215411558669e-05,
"loss": 0.1468,
"step": 3580
},
{
"epoch": 0.61,
"learning_rate": 1.8869964973730298e-05,
"loss": 0.277,
"step": 3585
},
{
"epoch": 0.62,
"learning_rate": 1.8867775831873908e-05,
"loss": 0.2693,
"step": 3590
},
{
"epoch": 0.62,
"learning_rate": 1.8865586690017514e-05,
"loss": 0.2667,
"step": 3595
},
{
"epoch": 0.62,
"learning_rate": 1.886339754816112e-05,
"loss": 0.4369,
"step": 3600
},
{
"epoch": 0.62,
"learning_rate": 1.886120840630473e-05,
"loss": 0.3818,
"step": 3605
},
{
"epoch": 0.62,
"learning_rate": 1.8859019264448337e-05,
"loss": 0.274,
"step": 3610
},
{
"epoch": 0.62,
"learning_rate": 1.8856830122591943e-05,
"loss": 0.1203,
"step": 3615
},
{
"epoch": 0.62,
"learning_rate": 1.8854640980735553e-05,
"loss": 0.1394,
"step": 3620
},
{
"epoch": 0.62,
"learning_rate": 1.885245183887916e-05,
"loss": 0.1707,
"step": 3625
},
{
"epoch": 0.62,
"learning_rate": 1.8850262697022766e-05,
"loss": 0.2372,
"step": 3630
},
{
"epoch": 0.62,
"learning_rate": 1.8848073555166376e-05,
"loss": 0.2037,
"step": 3635
},
{
"epoch": 0.62,
"learning_rate": 1.8845884413309986e-05,
"loss": 0.2806,
"step": 3640
},
{
"epoch": 0.62,
"learning_rate": 1.8843695271453592e-05,
"loss": 0.3228,
"step": 3645
},
{
"epoch": 0.63,
"learning_rate": 1.88415061295972e-05,
"loss": 0.5996,
"step": 3650
},
{
"epoch": 0.63,
"learning_rate": 1.883931698774081e-05,
"loss": 0.4127,
"step": 3655
},
{
"epoch": 0.63,
"learning_rate": 1.8837127845884415e-05,
"loss": 0.2125,
"step": 3660
},
{
"epoch": 0.63,
"learning_rate": 1.8834938704028025e-05,
"loss": 0.1208,
"step": 3665
},
{
"epoch": 0.63,
"learning_rate": 1.883274956217163e-05,
"loss": 0.1377,
"step": 3670
},
{
"epoch": 0.63,
"learning_rate": 1.8830560420315237e-05,
"loss": 0.1713,
"step": 3675
},
{
"epoch": 0.63,
"learning_rate": 1.8828371278458847e-05,
"loss": 0.1665,
"step": 3680
},
{
"epoch": 0.63,
"learning_rate": 1.8826182136602454e-05,
"loss": 0.2114,
"step": 3685
},
{
"epoch": 0.63,
"learning_rate": 1.882399299474606e-05,
"loss": 0.2506,
"step": 3690
},
{
"epoch": 0.63,
"learning_rate": 1.882180385288967e-05,
"loss": 0.3531,
"step": 3695
},
{
"epoch": 0.63,
"learning_rate": 1.8819614711033276e-05,
"loss": 0.4471,
"step": 3700
},
{
"epoch": 0.63,
"learning_rate": 1.8817425569176883e-05,
"loss": 0.3686,
"step": 3705
},
{
"epoch": 0.64,
"learning_rate": 1.8815236427320493e-05,
"loss": 0.3195,
"step": 3710
},
{
"epoch": 0.64,
"learning_rate": 1.88130472854641e-05,
"loss": 0.1353,
"step": 3715
},
{
"epoch": 0.64,
"learning_rate": 1.8810858143607706e-05,
"loss": 0.1753,
"step": 3720
},
{
"epoch": 0.64,
"learning_rate": 1.8808669001751315e-05,
"loss": 0.1398,
"step": 3725
},
{
"epoch": 0.64,
"learning_rate": 1.8806479859894922e-05,
"loss": 0.207,
"step": 3730
},
{
"epoch": 0.64,
"learning_rate": 1.880429071803853e-05,
"loss": 0.2271,
"step": 3735
},
{
"epoch": 0.64,
"learning_rate": 1.8802101576182138e-05,
"loss": 0.2495,
"step": 3740
},
{
"epoch": 0.64,
"learning_rate": 1.8799912434325748e-05,
"loss": 0.3315,
"step": 3745
},
{
"epoch": 0.64,
"learning_rate": 1.8797723292469354e-05,
"loss": 0.5553,
"step": 3750
},
{
"epoch": 0.64,
"learning_rate": 1.879553415061296e-05,
"loss": 0.4086,
"step": 3755
},
{
"epoch": 0.64,
"learning_rate": 1.879334500875657e-05,
"loss": 0.2226,
"step": 3760
},
{
"epoch": 0.65,
"learning_rate": 1.8791155866900177e-05,
"loss": 0.1356,
"step": 3765
},
{
"epoch": 0.65,
"learning_rate": 1.8788966725043783e-05,
"loss": 0.1158,
"step": 3770
},
{
"epoch": 0.65,
"learning_rate": 1.8786777583187393e-05,
"loss": 0.1887,
"step": 3775
},
{
"epoch": 0.65,
"learning_rate": 1.8784588441331e-05,
"loss": 0.197,
"step": 3780
},
{
"epoch": 0.65,
"learning_rate": 1.8782399299474606e-05,
"loss": 0.2012,
"step": 3785
},
{
"epoch": 0.65,
"learning_rate": 1.8780210157618216e-05,
"loss": 0.2624,
"step": 3790
},
{
"epoch": 0.65,
"learning_rate": 1.8778021015761822e-05,
"loss": 0.3297,
"step": 3795
},
{
"epoch": 0.65,
"learning_rate": 1.877583187390543e-05,
"loss": 0.4958,
"step": 3800
},
{
"epoch": 0.65,
"learning_rate": 1.877364273204904e-05,
"loss": 0.4359,
"step": 3805
},
{
"epoch": 0.65,
"learning_rate": 1.8771453590192645e-05,
"loss": 0.2487,
"step": 3810
},
{
"epoch": 0.65,
"learning_rate": 1.876926444833625e-05,
"loss": 0.1099,
"step": 3815
},
{
"epoch": 0.65,
"learning_rate": 1.876707530647986e-05,
"loss": 0.1422,
"step": 3820
},
{
"epoch": 0.66,
"learning_rate": 1.8764886164623468e-05,
"loss": 0.175,
"step": 3825
},
{
"epoch": 0.66,
"learning_rate": 1.8762697022767078e-05,
"loss": 0.2469,
"step": 3830
},
{
"epoch": 0.66,
"learning_rate": 1.8760507880910684e-05,
"loss": 0.2504,
"step": 3835
},
{
"epoch": 0.66,
"learning_rate": 1.8758318739054294e-05,
"loss": 0.2551,
"step": 3840
},
{
"epoch": 0.66,
"learning_rate": 1.87561295971979e-05,
"loss": 0.3085,
"step": 3845
},
{
"epoch": 0.66,
"learning_rate": 1.8753940455341507e-05,
"loss": 0.433,
"step": 3850
},
{
"epoch": 0.66,
"learning_rate": 1.8751751313485117e-05,
"loss": 0.3984,
"step": 3855
},
{
"epoch": 0.66,
"learning_rate": 1.8749562171628723e-05,
"loss": 0.2848,
"step": 3860
},
{
"epoch": 0.66,
"learning_rate": 1.874737302977233e-05,
"loss": 0.1222,
"step": 3865
},
{
"epoch": 0.66,
"learning_rate": 1.874518388791594e-05,
"loss": 0.2737,
"step": 3870
},
{
"epoch": 0.66,
"learning_rate": 1.8742994746059546e-05,
"loss": 0.156,
"step": 3875
},
{
"epoch": 0.66,
"learning_rate": 1.8740805604203156e-05,
"loss": 0.1961,
"step": 3880
},
{
"epoch": 0.67,
"learning_rate": 1.8738616462346762e-05,
"loss": 0.2205,
"step": 3885
},
{
"epoch": 0.67,
"learning_rate": 1.873642732049037e-05,
"loss": 0.2506,
"step": 3890
},
{
"epoch": 0.67,
"learning_rate": 1.8734238178633978e-05,
"loss": 0.3012,
"step": 3895
},
{
"epoch": 0.67,
"learning_rate": 1.8732049036777585e-05,
"loss": 0.464,
"step": 3900
},
{
"epoch": 0.67,
"learning_rate": 1.872985989492119e-05,
"loss": 0.3609,
"step": 3905
},
{
"epoch": 0.67,
"learning_rate": 1.87276707530648e-05,
"loss": 0.2539,
"step": 3910
},
{
"epoch": 0.67,
"learning_rate": 1.8725481611208407e-05,
"loss": 0.1313,
"step": 3915
},
{
"epoch": 0.67,
"learning_rate": 1.8723292469352014e-05,
"loss": 0.1494,
"step": 3920
},
{
"epoch": 0.67,
"learning_rate": 1.8721103327495624e-05,
"loss": 0.1587,
"step": 3925
},
{
"epoch": 0.67,
"learning_rate": 1.8718914185639233e-05,
"loss": 0.1675,
"step": 3930
},
{
"epoch": 0.67,
"learning_rate": 1.871672504378284e-05,
"loss": 0.2004,
"step": 3935
},
{
"epoch": 0.68,
"learning_rate": 1.8714535901926446e-05,
"loss": 0.2796,
"step": 3940
},
{
"epoch": 0.68,
"learning_rate": 1.8712346760070056e-05,
"loss": 0.2819,
"step": 3945
},
{
"epoch": 0.68,
"learning_rate": 1.8710157618213663e-05,
"loss": 0.5482,
"step": 3950
},
{
"epoch": 0.68,
"learning_rate": 1.870796847635727e-05,
"loss": 0.3684,
"step": 3955
},
{
"epoch": 0.68,
"learning_rate": 1.870577933450088e-05,
"loss": 0.2803,
"step": 3960
},
{
"epoch": 0.68,
"learning_rate": 1.8703590192644485e-05,
"loss": 0.1454,
"step": 3965
},
{
"epoch": 0.68,
"learning_rate": 1.8701401050788092e-05,
"loss": 0.1338,
"step": 3970
},
{
"epoch": 0.68,
"learning_rate": 1.86992119089317e-05,
"loss": 0.1544,
"step": 3975
},
{
"epoch": 0.68,
"learning_rate": 1.8697022767075308e-05,
"loss": 0.1506,
"step": 3980
},
{
"epoch": 0.68,
"learning_rate": 1.8694833625218914e-05,
"loss": 0.2438,
"step": 3985
},
{
"epoch": 0.68,
"learning_rate": 1.8692644483362524e-05,
"loss": 0.2494,
"step": 3990
},
{
"epoch": 0.68,
"learning_rate": 1.869045534150613e-05,
"loss": 0.3083,
"step": 3995
},
{
"epoch": 0.69,
"learning_rate": 1.8688266199649737e-05,
"loss": 0.4031,
"step": 4000
},
{
"epoch": 0.69,
"eval_loss": 0.09081415832042694,
"eval_runtime": 1458.9021,
"eval_samples_per_second": 14.223,
"eval_steps_per_second": 3.556,
"eval_wer": 0.418264082244592,
"step": 4000
},
{
"epoch": 0.69,
"learning_rate": 1.8686077057793347e-05,
"loss": 0.4179,
"step": 4005
},
{
"epoch": 0.69,
"learning_rate": 1.8683887915936953e-05,
"loss": 0.2825,
"step": 4010
},
{
"epoch": 0.69,
"learning_rate": 1.868169877408056e-05,
"loss": 0.0974,
"step": 4015
},
{
"epoch": 0.69,
"learning_rate": 1.867950963222417e-05,
"loss": 0.1454,
"step": 4020
},
{
"epoch": 0.69,
"learning_rate": 1.8677320490367776e-05,
"loss": 0.2178,
"step": 4025
},
{
"epoch": 0.69,
"learning_rate": 1.8675131348511386e-05,
"loss": 0.1735,
"step": 4030
},
{
"epoch": 0.69,
"learning_rate": 1.8672942206654992e-05,
"loss": 0.232,
"step": 4035
},
{
"epoch": 0.69,
"learning_rate": 1.8670753064798602e-05,
"loss": 0.2401,
"step": 4040
},
{
"epoch": 0.69,
"learning_rate": 1.866856392294221e-05,
"loss": 0.3915,
"step": 4045
},
{
"epoch": 0.69,
"learning_rate": 1.8666374781085815e-05,
"loss": 0.4997,
"step": 4050
},
{
"epoch": 0.69,
"learning_rate": 1.8664185639229425e-05,
"loss": 0.4202,
"step": 4055
},
{
"epoch": 0.7,
"learning_rate": 1.866199649737303e-05,
"loss": 0.2573,
"step": 4060
},
{
"epoch": 0.7,
"learning_rate": 1.8659807355516638e-05,
"loss": 0.1944,
"step": 4065
},
{
"epoch": 0.7,
"learning_rate": 1.8657618213660248e-05,
"loss": 0.1033,
"step": 4070
},
{
"epoch": 0.7,
"learning_rate": 1.8655429071803854e-05,
"loss": 0.1461,
"step": 4075
},
{
"epoch": 0.7,
"learning_rate": 1.865323992994746e-05,
"loss": 0.2328,
"step": 4080
},
{
"epoch": 0.7,
"learning_rate": 1.865105078809107e-05,
"loss": 0.273,
"step": 4085
},
{
"epoch": 0.7,
"learning_rate": 1.8648861646234677e-05,
"loss": 0.2329,
"step": 4090
},
{
"epoch": 0.7,
"learning_rate": 1.8646672504378283e-05,
"loss": 0.3154,
"step": 4095
},
{
"epoch": 0.7,
"learning_rate": 1.8644483362521893e-05,
"loss": 0.5075,
"step": 4100
},
{
"epoch": 0.7,
"learning_rate": 1.86422942206655e-05,
"loss": 0.3595,
"step": 4105
},
{
"epoch": 0.7,
"learning_rate": 1.864010507880911e-05,
"loss": 0.2888,
"step": 4110
},
{
"epoch": 0.71,
"learning_rate": 1.8637915936952716e-05,
"loss": 0.0926,
"step": 4115
},
{
"epoch": 0.71,
"learning_rate": 1.8635726795096322e-05,
"loss": 0.1322,
"step": 4120
},
{
"epoch": 0.71,
"learning_rate": 1.8633537653239932e-05,
"loss": 0.162,
"step": 4125
},
{
"epoch": 0.71,
"learning_rate": 1.8631348511383542e-05,
"loss": 0.1862,
"step": 4130
},
{
"epoch": 0.71,
"learning_rate": 1.8629159369527148e-05,
"loss": 0.2102,
"step": 4135
},
{
"epoch": 0.71,
"learning_rate": 1.8626970227670755e-05,
"loss": 0.2326,
"step": 4140
},
{
"epoch": 0.71,
"learning_rate": 1.8624781085814364e-05,
"loss": 0.3278,
"step": 4145
},
{
"epoch": 0.71,
"learning_rate": 1.862259194395797e-05,
"loss": 0.5288,
"step": 4150
},
{
"epoch": 0.71,
"learning_rate": 1.8620402802101577e-05,
"loss": 0.4043,
"step": 4155
},
{
"epoch": 0.71,
"learning_rate": 1.8618213660245187e-05,
"loss": 0.2834,
"step": 4160
},
{
"epoch": 0.71,
"learning_rate": 1.8616024518388794e-05,
"loss": 0.1023,
"step": 4165
},
{
"epoch": 0.71,
"learning_rate": 1.86138353765324e-05,
"loss": 0.1466,
"step": 4170
},
{
"epoch": 0.72,
"learning_rate": 1.861164623467601e-05,
"loss": 0.1499,
"step": 4175
},
{
"epoch": 0.72,
"learning_rate": 1.8609457092819616e-05,
"loss": 0.2086,
"step": 4180
},
{
"epoch": 0.72,
"learning_rate": 1.8607267950963223e-05,
"loss": 0.2081,
"step": 4185
},
{
"epoch": 0.72,
"learning_rate": 1.8605078809106833e-05,
"loss": 0.2178,
"step": 4190
},
{
"epoch": 0.72,
"learning_rate": 1.860288966725044e-05,
"loss": 0.3536,
"step": 4195
},
{
"epoch": 0.72,
"learning_rate": 1.8600700525394045e-05,
"loss": 0.4379,
"step": 4200
},
{
"epoch": 0.72,
"learning_rate": 1.8598511383537655e-05,
"loss": 0.3751,
"step": 4205
},
{
"epoch": 0.72,
"learning_rate": 1.859632224168126e-05,
"loss": 0.2542,
"step": 4210
},
{
"epoch": 0.72,
"learning_rate": 1.8594133099824868e-05,
"loss": 0.1095,
"step": 4215
},
{
"epoch": 0.72,
"learning_rate": 1.8591943957968478e-05,
"loss": 0.1686,
"step": 4220
},
{
"epoch": 0.72,
"learning_rate": 1.8589754816112088e-05,
"loss": 0.1401,
"step": 4225
},
{
"epoch": 0.72,
"learning_rate": 1.8587565674255694e-05,
"loss": 0.1672,
"step": 4230
},
{
"epoch": 0.73,
"learning_rate": 1.85853765323993e-05,
"loss": 0.1776,
"step": 4235
},
{
"epoch": 0.73,
"learning_rate": 1.858318739054291e-05,
"loss": 0.2071,
"step": 4240
},
{
"epoch": 0.73,
"learning_rate": 1.8580998248686517e-05,
"loss": 0.2768,
"step": 4245
},
{
"epoch": 0.73,
"learning_rate": 1.8578809106830123e-05,
"loss": 0.468,
"step": 4250
},
{
"epoch": 0.73,
"learning_rate": 1.8576619964973733e-05,
"loss": 0.4098,
"step": 4255
},
{
"epoch": 0.73,
"learning_rate": 1.857443082311734e-05,
"loss": 0.3181,
"step": 4260
},
{
"epoch": 0.73,
"learning_rate": 1.8572241681260946e-05,
"loss": 0.1042,
"step": 4265
},
{
"epoch": 0.73,
"learning_rate": 1.8570052539404556e-05,
"loss": 0.1185,
"step": 4270
},
{
"epoch": 0.73,
"learning_rate": 1.8567863397548162e-05,
"loss": 0.1549,
"step": 4275
},
{
"epoch": 0.73,
"learning_rate": 1.856567425569177e-05,
"loss": 0.1448,
"step": 4280
},
{
"epoch": 0.73,
"learning_rate": 1.856348511383538e-05,
"loss": 0.1801,
"step": 4285
},
{
"epoch": 0.74,
"learning_rate": 1.8561295971978985e-05,
"loss": 0.2856,
"step": 4290
},
{
"epoch": 0.74,
"learning_rate": 1.855910683012259e-05,
"loss": 0.314,
"step": 4295
},
{
"epoch": 0.74,
"learning_rate": 1.85569176882662e-05,
"loss": 0.5655,
"step": 4300
},
{
"epoch": 0.74,
"learning_rate": 1.8554728546409808e-05,
"loss": 0.4564,
"step": 4305
},
{
"epoch": 0.74,
"learning_rate": 1.8552539404553414e-05,
"loss": 0.2695,
"step": 4310
},
{
"epoch": 0.74,
"learning_rate": 1.8550350262697024e-05,
"loss": 0.1394,
"step": 4315
},
{
"epoch": 0.74,
"learning_rate": 1.854816112084063e-05,
"loss": 0.1439,
"step": 4320
},
{
"epoch": 0.74,
"learning_rate": 1.854597197898424e-05,
"loss": 0.1714,
"step": 4325
},
{
"epoch": 0.74,
"learning_rate": 1.854378283712785e-05,
"loss": 0.202,
"step": 4330
},
{
"epoch": 0.74,
"learning_rate": 1.8541593695271456e-05,
"loss": 0.2171,
"step": 4335
},
{
"epoch": 0.74,
"learning_rate": 1.8539404553415063e-05,
"loss": 0.2477,
"step": 4340
},
{
"epoch": 0.74,
"learning_rate": 1.8537215411558673e-05,
"loss": 0.2873,
"step": 4345
},
{
"epoch": 0.75,
"learning_rate": 1.853502626970228e-05,
"loss": 0.4059,
"step": 4350
},
{
"epoch": 0.75,
"learning_rate": 1.8532837127845886e-05,
"loss": 0.4592,
"step": 4355
},
{
"epoch": 0.75,
"learning_rate": 1.8530647985989495e-05,
"loss": 0.319,
"step": 4360
},
{
"epoch": 0.75,
"learning_rate": 1.8528458844133102e-05,
"loss": 0.0969,
"step": 4365
},
{
"epoch": 0.75,
"learning_rate": 1.8526269702276708e-05,
"loss": 0.1418,
"step": 4370
},
{
"epoch": 0.75,
"learning_rate": 1.8524080560420318e-05,
"loss": 0.1589,
"step": 4375
},
{
"epoch": 0.75,
"learning_rate": 1.8521891418563925e-05,
"loss": 0.1977,
"step": 4380
},
{
"epoch": 0.75,
"learning_rate": 1.851970227670753e-05,
"loss": 0.2217,
"step": 4385
},
{
"epoch": 0.75,
"learning_rate": 1.851751313485114e-05,
"loss": 0.2537,
"step": 4390
},
{
"epoch": 0.75,
"learning_rate": 1.8515323992994747e-05,
"loss": 0.2982,
"step": 4395
},
{
"epoch": 0.75,
"learning_rate": 1.8513134851138354e-05,
"loss": 0.4523,
"step": 4400
},
{
"epoch": 0.75,
"learning_rate": 1.8510945709281963e-05,
"loss": 0.3783,
"step": 4405
},
{
"epoch": 0.76,
"learning_rate": 1.850875656742557e-05,
"loss": 0.2072,
"step": 4410
},
{
"epoch": 0.76,
"learning_rate": 1.8506567425569176e-05,
"loss": 0.0875,
"step": 4415
},
{
"epoch": 0.76,
"learning_rate": 1.8504378283712786e-05,
"loss": 0.1281,
"step": 4420
},
{
"epoch": 0.76,
"learning_rate": 1.8502189141856396e-05,
"loss": 0.1564,
"step": 4425
},
{
"epoch": 0.76,
"learning_rate": 1.8500000000000002e-05,
"loss": 0.1761,
"step": 4430
},
{
"epoch": 0.76,
"learning_rate": 1.849781085814361e-05,
"loss": 0.219,
"step": 4435
},
{
"epoch": 0.76,
"learning_rate": 1.849562171628722e-05,
"loss": 0.276,
"step": 4440
},
{
"epoch": 0.76,
"learning_rate": 1.8493432574430825e-05,
"loss": 0.3416,
"step": 4445
},
{
"epoch": 0.76,
"learning_rate": 1.849124343257443e-05,
"loss": 0.4522,
"step": 4450
},
{
"epoch": 0.76,
"learning_rate": 1.848905429071804e-05,
"loss": 0.3271,
"step": 4455
},
{
"epoch": 0.76,
"learning_rate": 1.8486865148861648e-05,
"loss": 0.2506,
"step": 4460
},
{
"epoch": 0.77,
"learning_rate": 1.8484676007005254e-05,
"loss": 0.0979,
"step": 4465
},
{
"epoch": 0.77,
"learning_rate": 1.8482486865148864e-05,
"loss": 0.1518,
"step": 4470
},
{
"epoch": 0.77,
"learning_rate": 1.848029772329247e-05,
"loss": 0.1448,
"step": 4475
},
{
"epoch": 0.77,
"learning_rate": 1.8478108581436077e-05,
"loss": 0.1996,
"step": 4480
},
{
"epoch": 0.77,
"learning_rate": 1.8475919439579687e-05,
"loss": 0.1703,
"step": 4485
},
{
"epoch": 0.77,
"learning_rate": 1.8473730297723293e-05,
"loss": 0.2123,
"step": 4490
},
{
"epoch": 0.77,
"learning_rate": 1.84715411558669e-05,
"loss": 0.3168,
"step": 4495
},
{
"epoch": 0.77,
"learning_rate": 1.846935201401051e-05,
"loss": 0.4332,
"step": 4500
},
{
"epoch": 0.77,
"eval_loss": 0.08194659650325775,
"eval_runtime": 1659.6439,
"eval_samples_per_second": 12.503,
"eval_steps_per_second": 3.126,
"eval_wer": 0.35829406725208823,
"step": 4500
},
{
"epoch": 0.77,
"learning_rate": 1.8467162872154116e-05,
"loss": 0.3526,
"step": 4505
},
{
"epoch": 0.77,
"learning_rate": 1.8464973730297722e-05,
"loss": 0.2364,
"step": 4510
},
{
"epoch": 0.77,
"learning_rate": 1.8462784588441332e-05,
"loss": 0.1397,
"step": 4515
},
{
"epoch": 0.77,
"learning_rate": 1.8460595446584942e-05,
"loss": 0.1309,
"step": 4520
},
{
"epoch": 0.78,
"learning_rate": 1.845840630472855e-05,
"loss": 0.1971,
"step": 4525
},
{
"epoch": 0.78,
"learning_rate": 1.8456217162872155e-05,
"loss": 0.2176,
"step": 4530
},
{
"epoch": 0.78,
"learning_rate": 1.8454028021015765e-05,
"loss": 0.2123,
"step": 4535
},
{
"epoch": 0.78,
"learning_rate": 1.845183887915937e-05,
"loss": 0.2798,
"step": 4540
},
{
"epoch": 0.78,
"learning_rate": 1.8449649737302978e-05,
"loss": 0.3123,
"step": 4545
},
{
"epoch": 0.78,
"learning_rate": 1.8447460595446587e-05,
"loss": 0.42,
"step": 4550
},
{
"epoch": 0.78,
"learning_rate": 1.8445271453590194e-05,
"loss": 0.3659,
"step": 4555
},
{
"epoch": 0.78,
"learning_rate": 1.8443082311733804e-05,
"loss": 0.2704,
"step": 4560
},
{
"epoch": 0.78,
"learning_rate": 1.844089316987741e-05,
"loss": 0.1018,
"step": 4565
},
{
"epoch": 0.78,
"learning_rate": 1.8438704028021017e-05,
"loss": 0.1331,
"step": 4570
},
{
"epoch": 0.78,
"learning_rate": 1.8436514886164626e-05,
"loss": 0.1455,
"step": 4575
},
{
"epoch": 0.78,
"learning_rate": 1.8434325744308233e-05,
"loss": 0.2042,
"step": 4580
},
{
"epoch": 0.79,
"learning_rate": 1.843213660245184e-05,
"loss": 0.2272,
"step": 4585
},
{
"epoch": 0.79,
"learning_rate": 1.842994746059545e-05,
"loss": 0.2461,
"step": 4590
},
{
"epoch": 0.79,
"learning_rate": 1.8427758318739055e-05,
"loss": 0.2971,
"step": 4595
},
{
"epoch": 0.79,
"learning_rate": 1.8425569176882662e-05,
"loss": 0.4744,
"step": 4600
},
{
"epoch": 0.79,
"learning_rate": 1.8423380035026272e-05,
"loss": 0.3393,
"step": 4605
},
{
"epoch": 0.79,
"learning_rate": 1.8421190893169878e-05,
"loss": 0.2505,
"step": 4610
},
{
"epoch": 0.79,
"learning_rate": 1.8419001751313485e-05,
"loss": 0.1127,
"step": 4615
},
{
"epoch": 0.79,
"learning_rate": 1.8416812609457094e-05,
"loss": 0.1368,
"step": 4620
},
{
"epoch": 0.79,
"learning_rate": 1.8414623467600704e-05,
"loss": 0.1305,
"step": 4625
},
{
"epoch": 0.79,
"learning_rate": 1.841243432574431e-05,
"loss": 0.1457,
"step": 4630
},
{
"epoch": 0.79,
"learning_rate": 1.8410245183887917e-05,
"loss": 0.2298,
"step": 4635
},
{
"epoch": 0.8,
"learning_rate": 1.8408056042031527e-05,
"loss": 0.2438,
"step": 4640
},
{
"epoch": 0.8,
"learning_rate": 1.8405866900175133e-05,
"loss": 0.367,
"step": 4645
},
{
"epoch": 0.8,
"learning_rate": 1.840367775831874e-05,
"loss": 0.4247,
"step": 4650
},
{
"epoch": 0.8,
"learning_rate": 1.840148861646235e-05,
"loss": 0.3418,
"step": 4655
},
{
"epoch": 0.8,
"learning_rate": 1.8399299474605956e-05,
"loss": 0.2242,
"step": 4660
},
{
"epoch": 0.8,
"learning_rate": 1.8397110332749563e-05,
"loss": 0.0977,
"step": 4665
},
{
"epoch": 0.8,
"learning_rate": 1.8394921190893172e-05,
"loss": 0.1503,
"step": 4670
},
{
"epoch": 0.8,
"learning_rate": 1.839273204903678e-05,
"loss": 0.1398,
"step": 4675
},
{
"epoch": 0.8,
"learning_rate": 1.8390542907180385e-05,
"loss": 0.2352,
"step": 4680
},
{
"epoch": 0.8,
"learning_rate": 1.8388353765323995e-05,
"loss": 0.2031,
"step": 4685
},
{
"epoch": 0.8,
"learning_rate": 1.83861646234676e-05,
"loss": 0.2244,
"step": 4690
},
{
"epoch": 0.8,
"learning_rate": 1.8383975481611208e-05,
"loss": 0.2737,
"step": 4695
},
{
"epoch": 0.81,
"learning_rate": 1.8381786339754818e-05,
"loss": 0.4191,
"step": 4700
},
{
"epoch": 0.81,
"learning_rate": 1.8379597197898424e-05,
"loss": 0.3543,
"step": 4705
},
{
"epoch": 0.81,
"learning_rate": 1.837740805604203e-05,
"loss": 0.2448,
"step": 4710
},
{
"epoch": 0.81,
"learning_rate": 1.837521891418564e-05,
"loss": 0.1107,
"step": 4715
},
{
"epoch": 0.81,
"learning_rate": 1.837302977232925e-05,
"loss": 0.1249,
"step": 4720
},
{
"epoch": 0.81,
"learning_rate": 1.8370840630472857e-05,
"loss": 0.1464,
"step": 4725
},
{
"epoch": 0.81,
"learning_rate": 1.8368651488616463e-05,
"loss": 0.1829,
"step": 4730
},
{
"epoch": 0.81,
"learning_rate": 1.8366462346760073e-05,
"loss": 0.1933,
"step": 4735
},
{
"epoch": 0.81,
"learning_rate": 1.836427320490368e-05,
"loss": 0.2509,
"step": 4740
},
{
"epoch": 0.81,
"learning_rate": 1.8362084063047286e-05,
"loss": 0.2668,
"step": 4745
},
{
"epoch": 0.81,
"learning_rate": 1.8359894921190896e-05,
"loss": 0.479,
"step": 4750
},
{
"epoch": 0.81,
"learning_rate": 1.8357705779334502e-05,
"loss": 0.4035,
"step": 4755
},
{
"epoch": 0.82,
"learning_rate": 1.835551663747811e-05,
"loss": 0.2602,
"step": 4760
},
{
"epoch": 0.82,
"learning_rate": 1.835332749562172e-05,
"loss": 0.1103,
"step": 4765
},
{
"epoch": 0.82,
"learning_rate": 1.8351138353765325e-05,
"loss": 0.1307,
"step": 4770
},
{
"epoch": 0.82,
"learning_rate": 1.8348949211908935e-05,
"loss": 0.1544,
"step": 4775
},
{
"epoch": 0.82,
"learning_rate": 1.834676007005254e-05,
"loss": 0.1317,
"step": 4780
},
{
"epoch": 0.82,
"learning_rate": 1.8344570928196148e-05,
"loss": 0.2139,
"step": 4785
},
{
"epoch": 0.82,
"learning_rate": 1.8342381786339757e-05,
"loss": 0.2261,
"step": 4790
},
{
"epoch": 0.82,
"learning_rate": 1.8340192644483364e-05,
"loss": 0.2696,
"step": 4795
},
{
"epoch": 0.82,
"learning_rate": 1.833800350262697e-05,
"loss": 0.4188,
"step": 4800
},
{
"epoch": 0.82,
"learning_rate": 1.833581436077058e-05,
"loss": 0.3318,
"step": 4805
},
{
"epoch": 0.82,
"learning_rate": 1.8333625218914186e-05,
"loss": 0.2606,
"step": 4810
},
{
"epoch": 0.83,
"learning_rate": 1.8331436077057796e-05,
"loss": 0.103,
"step": 4815
},
{
"epoch": 0.83,
"learning_rate": 1.8329246935201403e-05,
"loss": 0.1346,
"step": 4820
},
{
"epoch": 0.83,
"learning_rate": 1.8327057793345013e-05,
"loss": 0.1226,
"step": 4825
},
{
"epoch": 0.83,
"learning_rate": 1.832486865148862e-05,
"loss": 0.1965,
"step": 4830
},
{
"epoch": 0.83,
"learning_rate": 1.8322679509632225e-05,
"loss": 0.2275,
"step": 4835
},
{
"epoch": 0.83,
"learning_rate": 1.8320490367775835e-05,
"loss": 0.2448,
"step": 4840
},
{
"epoch": 0.83,
"learning_rate": 1.831830122591944e-05,
"loss": 0.2616,
"step": 4845
},
{
"epoch": 0.83,
"learning_rate": 1.8316112084063048e-05,
"loss": 0.4799,
"step": 4850
},
{
"epoch": 0.83,
"learning_rate": 1.8313922942206658e-05,
"loss": 0.3476,
"step": 4855
},
{
"epoch": 0.83,
"learning_rate": 1.8311733800350264e-05,
"loss": 0.241,
"step": 4860
},
{
"epoch": 0.83,
"learning_rate": 1.830954465849387e-05,
"loss": 0.096,
"step": 4865
},
{
"epoch": 0.83,
"learning_rate": 1.830735551663748e-05,
"loss": 0.1357,
"step": 4870
},
{
"epoch": 0.84,
"learning_rate": 1.8305166374781087e-05,
"loss": 0.1547,
"step": 4875
},
{
"epoch": 0.84,
"learning_rate": 1.8302977232924694e-05,
"loss": 0.1348,
"step": 4880
},
{
"epoch": 0.84,
"learning_rate": 1.8300788091068303e-05,
"loss": 0.2907,
"step": 4885
},
{
"epoch": 0.84,
"learning_rate": 1.829859894921191e-05,
"loss": 0.2771,
"step": 4890
},
{
"epoch": 0.84,
"learning_rate": 1.8296409807355516e-05,
"loss": 0.2824,
"step": 4895
},
{
"epoch": 0.84,
"learning_rate": 1.8294220665499126e-05,
"loss": 0.462,
"step": 4900
},
{
"epoch": 0.84,
"learning_rate": 1.8292031523642732e-05,
"loss": 0.338,
"step": 4905
},
{
"epoch": 0.84,
"learning_rate": 1.828984238178634e-05,
"loss": 0.2117,
"step": 4910
},
{
"epoch": 0.84,
"learning_rate": 1.828765323992995e-05,
"loss": 0.1139,
"step": 4915
},
{
"epoch": 0.84,
"learning_rate": 1.828546409807356e-05,
"loss": 0.139,
"step": 4920
},
{
"epoch": 0.84,
"learning_rate": 1.8283274956217165e-05,
"loss": 0.1335,
"step": 4925
},
{
"epoch": 0.84,
"learning_rate": 1.828108581436077e-05,
"loss": 0.1872,
"step": 4930
},
{
"epoch": 0.85,
"learning_rate": 1.827889667250438e-05,
"loss": 0.1801,
"step": 4935
},
{
"epoch": 0.85,
"learning_rate": 1.8276707530647988e-05,
"loss": 0.2118,
"step": 4940
},
{
"epoch": 0.85,
"learning_rate": 1.8274518388791594e-05,
"loss": 0.2803,
"step": 4945
},
{
"epoch": 0.85,
"learning_rate": 1.8272329246935204e-05,
"loss": 0.4338,
"step": 4950
},
{
"epoch": 0.85,
"learning_rate": 1.827014010507881e-05,
"loss": 0.3406,
"step": 4955
},
{
"epoch": 0.85,
"learning_rate": 1.8267950963222417e-05,
"loss": 0.2522,
"step": 4960
},
{
"epoch": 0.85,
"learning_rate": 1.8265761821366027e-05,
"loss": 0.0804,
"step": 4965
},
{
"epoch": 0.85,
"learning_rate": 1.8263572679509633e-05,
"loss": 0.1033,
"step": 4970
},
{
"epoch": 0.85,
"learning_rate": 1.826138353765324e-05,
"loss": 0.1394,
"step": 4975
},
{
"epoch": 0.85,
"learning_rate": 1.825919439579685e-05,
"loss": 0.1191,
"step": 4980
},
{
"epoch": 0.85,
"learning_rate": 1.8257005253940456e-05,
"loss": 0.1947,
"step": 4985
},
{
"epoch": 0.86,
"learning_rate": 1.8254816112084062e-05,
"loss": 0.194,
"step": 4990
},
{
"epoch": 0.86,
"learning_rate": 1.8252626970227672e-05,
"loss": 0.2329,
"step": 4995
},
{
"epoch": 0.86,
"learning_rate": 1.825043782837128e-05,
"loss": 0.4045,
"step": 5000
},
{
"epoch": 0.86,
"eval_loss": 0.0839211717247963,
"eval_runtime": 1745.3682,
"eval_samples_per_second": 11.889,
"eval_steps_per_second": 2.972,
"eval_wer": 0.33275326622403084,
"step": 5000
},
{
"epoch": 0.86,
"learning_rate": 1.8248248686514888e-05,
"loss": 0.3367,
"step": 5005
},
{
"epoch": 0.86,
"learning_rate": 1.8246059544658495e-05,
"loss": 0.2498,
"step": 5010
},
{
"epoch": 0.86,
"learning_rate": 1.8243870402802105e-05,
"loss": 0.0952,
"step": 5015
},
{
"epoch": 0.86,
"learning_rate": 1.824168126094571e-05,
"loss": 0.1119,
"step": 5020
},
{
"epoch": 0.86,
"learning_rate": 1.823949211908932e-05,
"loss": 0.1133,
"step": 5025
},
{
"epoch": 0.86,
"learning_rate": 1.8237302977232927e-05,
"loss": 0.1574,
"step": 5030
},
{
"epoch": 0.86,
"learning_rate": 1.8235113835376534e-05,
"loss": 0.2018,
"step": 5035
},
{
"epoch": 0.86,
"learning_rate": 1.8232924693520144e-05,
"loss": 0.2064,
"step": 5040
},
{
"epoch": 0.86,
"learning_rate": 1.823073555166375e-05,
"loss": 0.2839,
"step": 5045
},
{
"epoch": 0.87,
"learning_rate": 1.8228546409807356e-05,
"loss": 0.4156,
"step": 5050
},
{
"epoch": 0.87,
"learning_rate": 1.8226357267950966e-05,
"loss": 0.3321,
"step": 5055
},
{
"epoch": 0.87,
"learning_rate": 1.8224168126094573e-05,
"loss": 0.2639,
"step": 5060
},
{
"epoch": 0.87,
"learning_rate": 1.822197898423818e-05,
"loss": 0.0911,
"step": 5065
},
{
"epoch": 0.87,
"learning_rate": 1.821978984238179e-05,
"loss": 0.144,
"step": 5070
},
{
"epoch": 0.87,
"learning_rate": 1.8217600700525395e-05,
"loss": 0.1596,
"step": 5075
},
{
"epoch": 0.87,
"learning_rate": 1.8215411558669002e-05,
"loss": 0.1685,
"step": 5080
},
{
"epoch": 0.87,
"learning_rate": 1.821322241681261e-05,
"loss": 0.2138,
"step": 5085
},
{
"epoch": 0.87,
"learning_rate": 1.8211033274956218e-05,
"loss": 0.2368,
"step": 5090
},
{
"epoch": 0.87,
"learning_rate": 1.8208844133099824e-05,
"loss": 0.2821,
"step": 5095
},
{
"epoch": 0.87,
"learning_rate": 1.8206654991243434e-05,
"loss": 0.4695,
"step": 5100
},
{
"epoch": 0.87,
"learning_rate": 1.820446584938704e-05,
"loss": 0.3143,
"step": 5105
},
{
"epoch": 0.88,
"learning_rate": 1.820227670753065e-05,
"loss": 0.2002,
"step": 5110
},
{
"epoch": 0.88,
"learning_rate": 1.8200087565674257e-05,
"loss": 0.0872,
"step": 5115
},
{
"epoch": 0.88,
"learning_rate": 1.8197898423817867e-05,
"loss": 0.1202,
"step": 5120
},
{
"epoch": 0.88,
"learning_rate": 1.8195709281961473e-05,
"loss": 0.1406,
"step": 5125
},
{
"epoch": 0.88,
"learning_rate": 1.819352014010508e-05,
"loss": 0.1624,
"step": 5130
},
{
"epoch": 0.88,
"learning_rate": 1.819133099824869e-05,
"loss": 0.1587,
"step": 5135
},
{
"epoch": 0.88,
"learning_rate": 1.8189141856392296e-05,
"loss": 0.1704,
"step": 5140
},
{
"epoch": 0.88,
"learning_rate": 1.8186952714535902e-05,
"loss": 0.3138,
"step": 5145
},
{
"epoch": 0.88,
"learning_rate": 1.8184763572679512e-05,
"loss": 0.4156,
"step": 5150
},
{
"epoch": 0.88,
"learning_rate": 1.818257443082312e-05,
"loss": 0.3135,
"step": 5155
},
{
"epoch": 0.88,
"learning_rate": 1.8180385288966725e-05,
"loss": 0.2094,
"step": 5160
},
{
"epoch": 0.89,
"learning_rate": 1.8178196147110335e-05,
"loss": 0.0802,
"step": 5165
},
{
"epoch": 0.89,
"learning_rate": 1.817600700525394e-05,
"loss": 0.1114,
"step": 5170
},
{
"epoch": 0.89,
"learning_rate": 1.8173817863397548e-05,
"loss": 0.1275,
"step": 5175
},
{
"epoch": 0.89,
"learning_rate": 1.8171628721541158e-05,
"loss": 0.1468,
"step": 5180
},
{
"epoch": 0.89,
"learning_rate": 1.8169439579684764e-05,
"loss": 0.2524,
"step": 5185
},
{
"epoch": 0.89,
"learning_rate": 1.816725043782837e-05,
"loss": 0.1925,
"step": 5190
},
{
"epoch": 0.89,
"learning_rate": 1.816506129597198e-05,
"loss": 0.2484,
"step": 5195
},
{
"epoch": 0.89,
"learning_rate": 1.8162872154115587e-05,
"loss": 0.4162,
"step": 5200
},
{
"epoch": 0.89,
"learning_rate": 1.8160683012259193e-05,
"loss": 0.3329,
"step": 5205
},
{
"epoch": 0.89,
"learning_rate": 1.8158493870402803e-05,
"loss": 0.2189,
"step": 5210
},
{
"epoch": 0.89,
"learning_rate": 1.8156304728546413e-05,
"loss": 0.1041,
"step": 5215
},
{
"epoch": 0.89,
"learning_rate": 1.815411558669002e-05,
"loss": 0.1104,
"step": 5220
},
{
"epoch": 0.9,
"learning_rate": 1.815192644483363e-05,
"loss": 0.1024,
"step": 5225
},
{
"epoch": 0.9,
"learning_rate": 1.8149737302977236e-05,
"loss": 0.1537,
"step": 5230
},
{
"epoch": 0.9,
"learning_rate": 1.8147548161120842e-05,
"loss": 0.1548,
"step": 5235
},
{
"epoch": 0.9,
"learning_rate": 1.8145359019264452e-05,
"loss": 0.2027,
"step": 5240
},
{
"epoch": 0.9,
"learning_rate": 1.8143169877408058e-05,
"loss": 0.2484,
"step": 5245
},
{
"epoch": 0.9,
"learning_rate": 1.8140980735551665e-05,
"loss": 0.3679,
"step": 5250
},
{
"epoch": 0.9,
"learning_rate": 1.8138791593695274e-05,
"loss": 0.3455,
"step": 5255
},
{
"epoch": 0.9,
"learning_rate": 1.813660245183888e-05,
"loss": 0.2438,
"step": 5260
},
{
"epoch": 0.9,
"learning_rate": 1.8134413309982487e-05,
"loss": 0.0783,
"step": 5265
},
{
"epoch": 0.9,
"learning_rate": 1.8132224168126097e-05,
"loss": 0.1011,
"step": 5270
},
{
"epoch": 0.9,
"learning_rate": 1.8130035026269704e-05,
"loss": 0.1118,
"step": 5275
},
{
"epoch": 0.9,
"learning_rate": 1.812784588441331e-05,
"loss": 0.2029,
"step": 5280
},
{
"epoch": 0.91,
"learning_rate": 1.812565674255692e-05,
"loss": 0.1785,
"step": 5285
},
{
"epoch": 0.91,
"learning_rate": 1.8123467600700526e-05,
"loss": 0.1993,
"step": 5290
},
{
"epoch": 0.91,
"learning_rate": 1.8121278458844133e-05,
"loss": 0.2923,
"step": 5295
},
{
"epoch": 0.91,
"learning_rate": 1.8119089316987743e-05,
"loss": 0.4213,
"step": 5300
},
{
"epoch": 0.91,
"learning_rate": 1.811690017513135e-05,
"loss": 0.3296,
"step": 5305
},
{
"epoch": 0.91,
"learning_rate": 1.811471103327496e-05,
"loss": 0.2241,
"step": 5310
},
{
"epoch": 0.91,
"learning_rate": 1.8112521891418565e-05,
"loss": 0.0837,
"step": 5315
},
{
"epoch": 0.91,
"learning_rate": 1.8110332749562175e-05,
"loss": 0.0946,
"step": 5320
},
{
"epoch": 0.91,
"learning_rate": 1.810814360770578e-05,
"loss": 0.1523,
"step": 5325
},
{
"epoch": 0.91,
"learning_rate": 1.8105954465849388e-05,
"loss": 0.1648,
"step": 5330
},
{
"epoch": 0.91,
"learning_rate": 1.8103765323992998e-05,
"loss": 0.1829,
"step": 5335
},
{
"epoch": 0.92,
"learning_rate": 1.8101576182136604e-05,
"loss": 0.1958,
"step": 5340
},
{
"epoch": 0.92,
"learning_rate": 1.809938704028021e-05,
"loss": 0.2725,
"step": 5345
},
{
"epoch": 0.92,
"learning_rate": 1.809719789842382e-05,
"loss": 0.3955,
"step": 5350
},
{
"epoch": 0.92,
"learning_rate": 1.8095008756567427e-05,
"loss": 0.3211,
"step": 5355
},
{
"epoch": 0.92,
"learning_rate": 1.8092819614711033e-05,
"loss": 0.2071,
"step": 5360
},
{
"epoch": 0.92,
"learning_rate": 1.8090630472854643e-05,
"loss": 0.1067,
"step": 5365
},
{
"epoch": 0.92,
"learning_rate": 1.808844133099825e-05,
"loss": 0.0993,
"step": 5370
},
{
"epoch": 0.92,
"learning_rate": 1.8086252189141856e-05,
"loss": 0.1648,
"step": 5375
},
{
"epoch": 0.92,
"learning_rate": 1.8084063047285466e-05,
"loss": 0.1649,
"step": 5380
},
{
"epoch": 0.92,
"learning_rate": 1.8081873905429072e-05,
"loss": 0.2296,
"step": 5385
},
{
"epoch": 0.92,
"learning_rate": 1.807968476357268e-05,
"loss": 0.1999,
"step": 5390
},
{
"epoch": 0.92,
"learning_rate": 1.807749562171629e-05,
"loss": 0.2402,
"step": 5395
},
{
"epoch": 0.93,
"learning_rate": 1.8075306479859895e-05,
"loss": 0.428,
"step": 5400
},
{
"epoch": 0.93,
"learning_rate": 1.8073117338003505e-05,
"loss": 0.3303,
"step": 5405
},
{
"epoch": 0.93,
"learning_rate": 1.807092819614711e-05,
"loss": 0.1896,
"step": 5410
},
{
"epoch": 0.93,
"learning_rate": 1.806873905429072e-05,
"loss": 0.0952,
"step": 5415
},
{
"epoch": 0.93,
"learning_rate": 1.8066549912434328e-05,
"loss": 0.1171,
"step": 5420
},
{
"epoch": 0.93,
"learning_rate": 1.8064360770577934e-05,
"loss": 0.1053,
"step": 5425
},
{
"epoch": 0.93,
"learning_rate": 1.8062171628721544e-05,
"loss": 0.1741,
"step": 5430
},
{
"epoch": 0.93,
"learning_rate": 1.805998248686515e-05,
"loss": 0.1631,
"step": 5435
},
{
"epoch": 0.93,
"learning_rate": 1.805779334500876e-05,
"loss": 0.1929,
"step": 5440
},
{
"epoch": 0.93,
"learning_rate": 1.8055604203152366e-05,
"loss": 0.2728,
"step": 5445
},
{
"epoch": 0.93,
"learning_rate": 1.8053415061295973e-05,
"loss": 0.4162,
"step": 5450
},
{
"epoch": 0.93,
"learning_rate": 1.8051225919439583e-05,
"loss": 0.302,
"step": 5455
},
{
"epoch": 0.94,
"learning_rate": 1.804903677758319e-05,
"loss": 0.1883,
"step": 5460
},
{
"epoch": 0.94,
"learning_rate": 1.8046847635726796e-05,
"loss": 0.0763,
"step": 5465
},
{
"epoch": 0.94,
"learning_rate": 1.8044658493870405e-05,
"loss": 0.1196,
"step": 5470
},
{
"epoch": 0.94,
"learning_rate": 1.8042469352014012e-05,
"loss": 0.1162,
"step": 5475
},
{
"epoch": 0.94,
"learning_rate": 1.804028021015762e-05,
"loss": 0.1665,
"step": 5480
},
{
"epoch": 0.94,
"learning_rate": 1.8038091068301228e-05,
"loss": 0.1563,
"step": 5485
},
{
"epoch": 0.94,
"learning_rate": 1.8035901926444835e-05,
"loss": 0.209,
"step": 5490
},
{
"epoch": 0.94,
"learning_rate": 1.803371278458844e-05,
"loss": 0.2812,
"step": 5495
},
{
"epoch": 0.94,
"learning_rate": 1.803152364273205e-05,
"loss": 0.4587,
"step": 5500
},
{
"epoch": 0.94,
"eval_loss": 0.07299930602312088,
"eval_runtime": 1565.9914,
"eval_samples_per_second": 13.25,
"eval_steps_per_second": 3.313,
"eval_wer": 0.32049153994431356,
"step": 5500
},
{
"epoch": 0.94,
"learning_rate": 1.802933450087566e-05,
"loss": 0.2905,
"step": 5505
},
{
"epoch": 0.94,
"learning_rate": 1.8027145359019267e-05,
"loss": 0.2116,
"step": 5510
},
{
"epoch": 0.95,
"learning_rate": 1.8024956217162874e-05,
"loss": 0.0778,
"step": 5515
},
{
"epoch": 0.95,
"learning_rate": 1.8022767075306483e-05,
"loss": 0.0997,
"step": 5520
},
{
"epoch": 0.95,
"learning_rate": 1.802057793345009e-05,
"loss": 0.1063,
"step": 5525
},
{
"epoch": 0.95,
"learning_rate": 1.8018388791593696e-05,
"loss": 0.1313,
"step": 5530
},
{
"epoch": 0.95,
"learning_rate": 1.8016199649737306e-05,
"loss": 0.2153,
"step": 5535
},
{
"epoch": 0.95,
"learning_rate": 1.8014010507880912e-05,
"loss": 0.1789,
"step": 5540
},
{
"epoch": 0.95,
"learning_rate": 1.801182136602452e-05,
"loss": 0.2895,
"step": 5545
},
{
"epoch": 0.95,
"learning_rate": 1.800963222416813e-05,
"loss": 0.4241,
"step": 5550
},
{
"epoch": 0.95,
"learning_rate": 1.8007443082311735e-05,
"loss": 0.3173,
"step": 5555
},
{
"epoch": 0.95,
"learning_rate": 1.800525394045534e-05,
"loss": 0.2008,
"step": 5560
},
{
"epoch": 0.95,
"learning_rate": 1.800306479859895e-05,
"loss": 0.1065,
"step": 5565
},
{
"epoch": 0.95,
"learning_rate": 1.8000875656742558e-05,
"loss": 0.1078,
"step": 5570
},
{
"epoch": 0.96,
"learning_rate": 1.7998686514886164e-05,
"loss": 0.1411,
"step": 5575
},
{
"epoch": 0.96,
"learning_rate": 1.7996497373029774e-05,
"loss": 0.143,
"step": 5580
},
{
"epoch": 0.96,
"learning_rate": 1.799430823117338e-05,
"loss": 0.1944,
"step": 5585
},
{
"epoch": 0.96,
"learning_rate": 1.7992119089316987e-05,
"loss": 0.2101,
"step": 5590
},
{
"epoch": 0.96,
"learning_rate": 1.7989929947460597e-05,
"loss": 0.2429,
"step": 5595
},
{
"epoch": 0.96,
"learning_rate": 1.7987740805604203e-05,
"loss": 0.3896,
"step": 5600
},
{
"epoch": 0.96,
"learning_rate": 1.7985551663747813e-05,
"loss": 0.2869,
"step": 5605
},
{
"epoch": 0.96,
"learning_rate": 1.798336252189142e-05,
"loss": 0.1854,
"step": 5610
},
{
"epoch": 0.96,
"learning_rate": 1.798117338003503e-05,
"loss": 0.0877,
"step": 5615
},
{
"epoch": 0.96,
"learning_rate": 1.7978984238178636e-05,
"loss": 0.1073,
"step": 5620
},
{
"epoch": 0.96,
"learning_rate": 1.7976795096322242e-05,
"loss": 0.131,
"step": 5625
},
{
"epoch": 0.96,
"learning_rate": 1.7974605954465852e-05,
"loss": 0.1769,
"step": 5630
},
{
"epoch": 0.97,
"learning_rate": 1.797241681260946e-05,
"loss": 0.184,
"step": 5635
},
{
"epoch": 0.97,
"learning_rate": 1.7970227670753065e-05,
"loss": 0.1882,
"step": 5640
},
{
"epoch": 0.97,
"learning_rate": 1.7968038528896675e-05,
"loss": 0.2627,
"step": 5645
},
{
"epoch": 0.97,
"learning_rate": 1.796584938704028e-05,
"loss": 0.4749,
"step": 5650
},
{
"epoch": 0.97,
"learning_rate": 1.7963660245183888e-05,
"loss": 0.308,
"step": 5655
},
{
"epoch": 0.97,
"learning_rate": 1.7961471103327497e-05,
"loss": 0.2261,
"step": 5660
},
{
"epoch": 0.97,
"learning_rate": 1.7959281961471104e-05,
"loss": 0.0864,
"step": 5665
},
{
"epoch": 0.97,
"learning_rate": 1.7957092819614714e-05,
"loss": 0.0929,
"step": 5670
},
{
"epoch": 0.97,
"learning_rate": 1.795490367775832e-05,
"loss": 0.1367,
"step": 5675
},
{
"epoch": 0.97,
"learning_rate": 1.7952714535901927e-05,
"loss": 0.1342,
"step": 5680
},
{
"epoch": 0.97,
"learning_rate": 1.7950525394045536e-05,
"loss": 0.1793,
"step": 5685
},
{
"epoch": 0.98,
"learning_rate": 1.7948336252189143e-05,
"loss": 0.2569,
"step": 5690
},
{
"epoch": 0.98,
"learning_rate": 1.794614711033275e-05,
"loss": 0.2253,
"step": 5695
},
{
"epoch": 0.98,
"learning_rate": 1.794395796847636e-05,
"loss": 0.4123,
"step": 5700
},
{
"epoch": 0.98,
"learning_rate": 1.794176882661997e-05,
"loss": 0.3305,
"step": 5705
},
{
"epoch": 0.98,
"learning_rate": 1.7939579684763575e-05,
"loss": 0.1793,
"step": 5710
},
{
"epoch": 0.98,
"learning_rate": 1.7937390542907182e-05,
"loss": 0.0935,
"step": 5715
},
{
"epoch": 0.98,
"learning_rate": 1.793520140105079e-05,
"loss": 0.0999,
"step": 5720
},
{
"epoch": 0.98,
"learning_rate": 1.7933012259194398e-05,
"loss": 0.1481,
"step": 5725
},
{
"epoch": 0.98,
"learning_rate": 1.7930823117338004e-05,
"loss": 0.1595,
"step": 5730
},
{
"epoch": 0.98,
"learning_rate": 1.7928633975481614e-05,
"loss": 0.1855,
"step": 5735
},
{
"epoch": 0.98,
"learning_rate": 1.792644483362522e-05,
"loss": 0.1865,
"step": 5740
},
{
"epoch": 0.98,
"learning_rate": 1.7924255691768827e-05,
"loss": 0.2998,
"step": 5745
},
{
"epoch": 0.99,
"learning_rate": 1.7922066549912437e-05,
"loss": 0.5011,
"step": 5750
},
{
"epoch": 0.99,
"learning_rate": 1.7919877408056043e-05,
"loss": 0.3,
"step": 5755
},
{
"epoch": 0.99,
"learning_rate": 1.791768826619965e-05,
"loss": 0.1897,
"step": 5760
},
{
"epoch": 0.99,
"learning_rate": 1.791549912434326e-05,
"loss": 0.0815,
"step": 5765
},
{
"epoch": 0.99,
"learning_rate": 1.7913309982486866e-05,
"loss": 0.116,
"step": 5770
},
{
"epoch": 0.99,
"learning_rate": 1.7911120840630473e-05,
"loss": 0.1409,
"step": 5775
},
{
"epoch": 0.99,
"learning_rate": 1.7908931698774082e-05,
"loss": 0.1524,
"step": 5780
},
{
"epoch": 0.99,
"learning_rate": 1.790674255691769e-05,
"loss": 0.1425,
"step": 5785
},
{
"epoch": 0.99,
"learning_rate": 1.7904553415061295e-05,
"loss": 0.2782,
"step": 5790
},
{
"epoch": 0.99,
"learning_rate": 1.7902364273204905e-05,
"loss": 0.2795,
"step": 5795
},
{
"epoch": 0.99,
"learning_rate": 1.7900175131348515e-05,
"loss": 0.5082,
"step": 5800
},
{
"epoch": 0.99,
"learning_rate": 1.789798598949212e-05,
"loss": 0.3098,
"step": 5805
},
{
"epoch": 1.0,
"learning_rate": 1.7895796847635728e-05,
"loss": 0.1319,
"step": 5810
},
{
"epoch": 1.0,
"learning_rate": 1.7893607705779338e-05,
"loss": 0.1048,
"step": 5815
},
{
"epoch": 1.0,
"learning_rate": 1.7891418563922944e-05,
"loss": 0.1553,
"step": 5820
},
{
"epoch": 1.0,
"learning_rate": 1.788922942206655e-05,
"loss": 0.2003,
"step": 5825
},
{
"epoch": 1.0,
"learning_rate": 1.788704028021016e-05,
"loss": 0.1832,
"step": 5830
},
{
"epoch": 1.0,
"learning_rate": 1.7884851138353767e-05,
"loss": 0.3404,
"step": 5835
},
{
"epoch": 1.0,
"learning_rate": 1.7882661996497373e-05,
"loss": 0.4007,
"step": 5840
},
{
"epoch": 1.0,
"learning_rate": 1.7880472854640983e-05,
"loss": 0.2453,
"step": 5845
},
{
"epoch": 1.0,
"learning_rate": 1.787828371278459e-05,
"loss": 0.0876,
"step": 5850
},
{
"epoch": 1.0,
"learning_rate": 1.7876094570928196e-05,
"loss": 0.0928,
"step": 5855
},
{
"epoch": 1.0,
"learning_rate": 1.7873905429071806e-05,
"loss": 0.1225,
"step": 5860
},
{
"epoch": 1.01,
"learning_rate": 1.7871716287215412e-05,
"loss": 0.1576,
"step": 5865
},
{
"epoch": 1.01,
"learning_rate": 1.786952714535902e-05,
"loss": 0.132,
"step": 5870
},
{
"epoch": 1.01,
"learning_rate": 1.786733800350263e-05,
"loss": 0.2191,
"step": 5875
},
{
"epoch": 1.01,
"learning_rate": 1.7865148861646235e-05,
"loss": 0.3289,
"step": 5880
},
{
"epoch": 1.01,
"learning_rate": 1.7862959719789845e-05,
"loss": 0.4885,
"step": 5885
},
{
"epoch": 1.01,
"learning_rate": 1.786077057793345e-05,
"loss": 0.3144,
"step": 5890
},
{
"epoch": 1.01,
"learning_rate": 1.7858581436077058e-05,
"loss": 0.1863,
"step": 5895
},
{
"epoch": 1.01,
"learning_rate": 1.7856392294220667e-05,
"loss": 0.0923,
"step": 5900
},
{
"epoch": 1.01,
"learning_rate": 1.7854203152364277e-05,
"loss": 0.0865,
"step": 5905
},
{
"epoch": 1.01,
"learning_rate": 1.7852014010507884e-05,
"loss": 0.0937,
"step": 5910
},
{
"epoch": 1.01,
"learning_rate": 1.784982486865149e-05,
"loss": 0.1737,
"step": 5915
},
{
"epoch": 1.01,
"learning_rate": 1.78476357267951e-05,
"loss": 0.1861,
"step": 5920
},
{
"epoch": 1.02,
"learning_rate": 1.7845446584938706e-05,
"loss": 0.2089,
"step": 5925
},
{
"epoch": 1.02,
"learning_rate": 1.7843257443082313e-05,
"loss": 0.261,
"step": 5930
},
{
"epoch": 1.02,
"learning_rate": 1.7841068301225923e-05,
"loss": 0.4025,
"step": 5935
},
{
"epoch": 1.02,
"learning_rate": 1.783887915936953e-05,
"loss": 0.2874,
"step": 5940
},
{
"epoch": 1.02,
"learning_rate": 1.7836690017513135e-05,
"loss": 0.1972,
"step": 5945
},
{
"epoch": 1.02,
"learning_rate": 1.7834500875656745e-05,
"loss": 0.0723,
"step": 5950
},
{
"epoch": 1.02,
"learning_rate": 1.7832311733800352e-05,
"loss": 0.0803,
"step": 5955
},
{
"epoch": 1.02,
"learning_rate": 1.7830122591943958e-05,
"loss": 0.1319,
"step": 5960
},
{
"epoch": 1.02,
"learning_rate": 1.7827933450087568e-05,
"loss": 0.1683,
"step": 5965
},
{
"epoch": 1.02,
"learning_rate": 1.7825744308231174e-05,
"loss": 0.1891,
"step": 5970
},
{
"epoch": 1.02,
"learning_rate": 1.782355516637478e-05,
"loss": 0.2245,
"step": 5975
},
{
"epoch": 1.02,
"learning_rate": 1.782136602451839e-05,
"loss": 0.2639,
"step": 5980
},
{
"epoch": 1.03,
"learning_rate": 1.7819176882661997e-05,
"loss": 0.3804,
"step": 5985
},
{
"epoch": 1.03,
"learning_rate": 1.7816987740805604e-05,
"loss": 0.3335,
"step": 5990
},
{
"epoch": 1.03,
"learning_rate": 1.7814798598949213e-05,
"loss": 0.2562,
"step": 5995
},
{
"epoch": 1.03,
"learning_rate": 1.7812609457092823e-05,
"loss": 0.0782,
"step": 6000
},
{
"epoch": 1.03,
"eval_loss": 0.07903540134429932,
"eval_runtime": 2315.0176,
"eval_samples_per_second": 8.963,
"eval_steps_per_second": 2.241,
"eval_wer": 0.2913364746198329,
"step": 6000
},
{
"epoch": 1.03,
"learning_rate": 1.781042031523643e-05,
"loss": 0.0858,
"step": 6005
},
{
"epoch": 1.03,
"learning_rate": 1.7808231173380036e-05,
"loss": 0.118,
"step": 6010
},
{
"epoch": 1.03,
"learning_rate": 1.7806042031523646e-05,
"loss": 0.121,
"step": 6015
},
{
"epoch": 1.03,
"learning_rate": 1.7803852889667252e-05,
"loss": 0.1664,
"step": 6020
},
{
"epoch": 1.03,
"learning_rate": 1.780166374781086e-05,
"loss": 0.1872,
"step": 6025
},
{
"epoch": 1.03,
"learning_rate": 1.779947460595447e-05,
"loss": 0.2711,
"step": 6030
},
{
"epoch": 1.03,
"learning_rate": 1.7797285464098075e-05,
"loss": 0.4361,
"step": 6035
},
{
"epoch": 1.04,
"learning_rate": 1.779509632224168e-05,
"loss": 0.2879,
"step": 6040
},
{
"epoch": 1.04,
"learning_rate": 1.779290718038529e-05,
"loss": 0.2426,
"step": 6045
},
{
"epoch": 1.04,
"learning_rate": 1.7790718038528898e-05,
"loss": 0.0981,
"step": 6050
},
{
"epoch": 1.04,
"learning_rate": 1.7788528896672504e-05,
"loss": 0.1269,
"step": 6055
},
{
"epoch": 1.04,
"learning_rate": 1.7786339754816114e-05,
"loss": 0.1434,
"step": 6060
},
{
"epoch": 1.04,
"learning_rate": 1.778415061295972e-05,
"loss": 0.1619,
"step": 6065
},
{
"epoch": 1.04,
"learning_rate": 1.7781961471103327e-05,
"loss": 0.1303,
"step": 6070
},
{
"epoch": 1.04,
"learning_rate": 1.7779772329246937e-05,
"loss": 0.1763,
"step": 6075
},
{
"epoch": 1.04,
"learning_rate": 1.7777583187390543e-05,
"loss": 0.2016,
"step": 6080
},
{
"epoch": 1.04,
"learning_rate": 1.777539404553415e-05,
"loss": 0.37,
"step": 6085
},
{
"epoch": 1.04,
"learning_rate": 1.777320490367776e-05,
"loss": 0.3281,
"step": 6090
},
{
"epoch": 1.04,
"learning_rate": 1.777101576182137e-05,
"loss": 0.1914,
"step": 6095
},
{
"epoch": 1.05,
"learning_rate": 1.7768826619964976e-05,
"loss": 0.0585,
"step": 6100
},
{
"epoch": 1.05,
"learning_rate": 1.7766637478108582e-05,
"loss": 0.1341,
"step": 6105
},
{
"epoch": 1.05,
"learning_rate": 1.7764448336252192e-05,
"loss": 0.1277,
"step": 6110
},
{
"epoch": 1.05,
"learning_rate": 1.77622591943958e-05,
"loss": 0.1888,
"step": 6115
},
{
"epoch": 1.05,
"learning_rate": 1.7760070052539408e-05,
"loss": 0.1709,
"step": 6120
},
{
"epoch": 1.05,
"learning_rate": 1.7757880910683015e-05,
"loss": 0.2112,
"step": 6125
},
{
"epoch": 1.05,
"learning_rate": 1.775569176882662e-05,
"loss": 0.2742,
"step": 6130
},
{
"epoch": 1.05,
"learning_rate": 1.775350262697023e-05,
"loss": 0.4163,
"step": 6135
},
{
"epoch": 1.05,
"learning_rate": 1.7751313485113837e-05,
"loss": 0.3179,
"step": 6140
},
{
"epoch": 1.05,
"learning_rate": 1.7749124343257444e-05,
"loss": 0.2413,
"step": 6145
},
{
"epoch": 1.05,
"learning_rate": 1.7746935201401054e-05,
"loss": 0.0621,
"step": 6150
},
{
"epoch": 1.05,
"learning_rate": 1.774474605954466e-05,
"loss": 0.1146,
"step": 6155
},
{
"epoch": 1.06,
"learning_rate": 1.7742556917688266e-05,
"loss": 0.0972,
"step": 6160
},
{
"epoch": 1.06,
"learning_rate": 1.7740367775831876e-05,
"loss": 0.1094,
"step": 6165
},
{
"epoch": 1.06,
"learning_rate": 1.7738178633975483e-05,
"loss": 0.1713,
"step": 6170
},
{
"epoch": 1.06,
"learning_rate": 1.773598949211909e-05,
"loss": 0.1844,
"step": 6175
},
{
"epoch": 1.06,
"learning_rate": 1.77338003502627e-05,
"loss": 0.2474,
"step": 6180
},
{
"epoch": 1.06,
"learning_rate": 1.7731611208406305e-05,
"loss": 0.4051,
"step": 6185
},
{
"epoch": 1.06,
"learning_rate": 1.7729422066549912e-05,
"loss": 0.2949,
"step": 6190
},
{
"epoch": 1.06,
"learning_rate": 1.772723292469352e-05,
"loss": 0.1988,
"step": 6195
},
{
"epoch": 1.06,
"learning_rate": 1.772504378283713e-05,
"loss": 0.0982,
"step": 6200
},
{
"epoch": 1.06,
"learning_rate": 1.7722854640980738e-05,
"loss": 0.084,
"step": 6205
},
{
"epoch": 1.06,
"learning_rate": 1.7720665499124344e-05,
"loss": 0.1077,
"step": 6210
},
{
"epoch": 1.07,
"learning_rate": 1.7718476357267954e-05,
"loss": 0.1275,
"step": 6215
},
{
"epoch": 1.07,
"learning_rate": 1.771628721541156e-05,
"loss": 0.1915,
"step": 6220
},
{
"epoch": 1.07,
"learning_rate": 1.7714098073555167e-05,
"loss": 0.1559,
"step": 6225
},
{
"epoch": 1.07,
"learning_rate": 1.7711908931698777e-05,
"loss": 0.2801,
"step": 6230
},
{
"epoch": 1.07,
"learning_rate": 1.7709719789842383e-05,
"loss": 0.4151,
"step": 6235
},
{
"epoch": 1.07,
"learning_rate": 1.770753064798599e-05,
"loss": 0.2865,
"step": 6240
},
{
"epoch": 1.07,
"learning_rate": 1.77053415061296e-05,
"loss": 0.192,
"step": 6245
},
{
"epoch": 1.07,
"learning_rate": 1.7703152364273206e-05,
"loss": 0.0771,
"step": 6250
},
{
"epoch": 1.07,
"learning_rate": 1.7700963222416812e-05,
"loss": 0.0881,
"step": 6255
},
{
"epoch": 1.07,
"learning_rate": 1.7698774080560422e-05,
"loss": 0.1123,
"step": 6260
},
{
"epoch": 1.07,
"learning_rate": 1.769658493870403e-05,
"loss": 0.1752,
"step": 6265
},
{
"epoch": 1.07,
"learning_rate": 1.7694395796847635e-05,
"loss": 0.2028,
"step": 6270
},
{
"epoch": 1.08,
"learning_rate": 1.7692206654991245e-05,
"loss": 0.1812,
"step": 6275
},
{
"epoch": 1.08,
"learning_rate": 1.769001751313485e-05,
"loss": 0.3254,
"step": 6280
},
{
"epoch": 1.08,
"learning_rate": 1.7687828371278458e-05,
"loss": 0.4148,
"step": 6285
},
{
"epoch": 1.08,
"learning_rate": 1.7685639229422068e-05,
"loss": 0.2792,
"step": 6290
},
{
"epoch": 1.08,
"learning_rate": 1.7683450087565677e-05,
"loss": 0.2164,
"step": 6295
},
{
"epoch": 1.08,
"learning_rate": 1.7681260945709284e-05,
"loss": 0.0654,
"step": 6300
},
{
"epoch": 1.08,
"learning_rate": 1.767907180385289e-05,
"loss": 0.0883,
"step": 6305
},
{
"epoch": 1.08,
"learning_rate": 1.76768826619965e-05,
"loss": 0.1371,
"step": 6310
},
{
"epoch": 1.08,
"learning_rate": 1.7674693520140107e-05,
"loss": 0.1435,
"step": 6315
},
{
"epoch": 1.08,
"learning_rate": 1.7672504378283713e-05,
"loss": 0.1577,
"step": 6320
},
{
"epoch": 1.08,
"learning_rate": 1.7670315236427323e-05,
"loss": 0.2004,
"step": 6325
},
{
"epoch": 1.08,
"learning_rate": 1.766812609457093e-05,
"loss": 0.2392,
"step": 6330
},
{
"epoch": 1.09,
"learning_rate": 1.766593695271454e-05,
"loss": 0.3124,
"step": 6335
},
{
"epoch": 1.09,
"learning_rate": 1.7663747810858146e-05,
"loss": 0.3016,
"step": 6340
},
{
"epoch": 1.09,
"learning_rate": 1.7661558669001752e-05,
"loss": 0.191,
"step": 6345
},
{
"epoch": 1.09,
"learning_rate": 1.7659369527145362e-05,
"loss": 0.0758,
"step": 6350
},
{
"epoch": 1.09,
"learning_rate": 1.7657180385288968e-05,
"loss": 0.0998,
"step": 6355
},
{
"epoch": 1.09,
"learning_rate": 1.7654991243432575e-05,
"loss": 0.1236,
"step": 6360
},
{
"epoch": 1.09,
"learning_rate": 1.7652802101576185e-05,
"loss": 0.1333,
"step": 6365
},
{
"epoch": 1.09,
"learning_rate": 1.765061295971979e-05,
"loss": 0.1703,
"step": 6370
},
{
"epoch": 1.09,
"learning_rate": 1.7648423817863397e-05,
"loss": 0.1676,
"step": 6375
},
{
"epoch": 1.09,
"learning_rate": 1.7646234676007007e-05,
"loss": 0.2243,
"step": 6380
},
{
"epoch": 1.09,
"learning_rate": 1.7644045534150614e-05,
"loss": 0.3975,
"step": 6385
},
{
"epoch": 1.1,
"learning_rate": 1.7641856392294223e-05,
"loss": 0.2579,
"step": 6390
},
{
"epoch": 1.1,
"learning_rate": 1.763966725043783e-05,
"loss": 0.1995,
"step": 6395
},
{
"epoch": 1.1,
"learning_rate": 1.763747810858144e-05,
"loss": 0.0853,
"step": 6400
},
{
"epoch": 1.1,
"learning_rate": 1.7635288966725046e-05,
"loss": 0.1125,
"step": 6405
},
{
"epoch": 1.1,
"learning_rate": 1.7633099824868653e-05,
"loss": 0.1067,
"step": 6410
},
{
"epoch": 1.1,
"learning_rate": 1.7630910683012262e-05,
"loss": 0.1362,
"step": 6415
},
{
"epoch": 1.1,
"learning_rate": 1.762872154115587e-05,
"loss": 0.1589,
"step": 6420
},
{
"epoch": 1.1,
"learning_rate": 1.7626532399299475e-05,
"loss": 0.1591,
"step": 6425
},
{
"epoch": 1.1,
"learning_rate": 1.7624343257443085e-05,
"loss": 0.4543,
"step": 6430
},
{
"epoch": 1.1,
"learning_rate": 1.762215411558669e-05,
"loss": 0.3441,
"step": 6435
},
{
"epoch": 1.1,
"learning_rate": 1.7619964973730298e-05,
"loss": 0.2826,
"step": 6440
},
{
"epoch": 1.1,
"learning_rate": 1.7617775831873908e-05,
"loss": 0.1319,
"step": 6445
},
{
"epoch": 1.11,
"learning_rate": 1.7615586690017514e-05,
"loss": 0.082,
"step": 6450
},
{
"epoch": 1.11,
"learning_rate": 1.761339754816112e-05,
"loss": 0.0804,
"step": 6455
},
{
"epoch": 1.11,
"learning_rate": 1.761120840630473e-05,
"loss": 0.095,
"step": 6460
},
{
"epoch": 1.11,
"learning_rate": 1.7609019264448337e-05,
"loss": 0.1319,
"step": 6465
},
{
"epoch": 1.11,
"learning_rate": 1.7606830122591943e-05,
"loss": 0.1588,
"step": 6470
},
{
"epoch": 1.11,
"learning_rate": 1.7604640980735553e-05,
"loss": 0.1866,
"step": 6475
},
{
"epoch": 1.11,
"learning_rate": 1.760245183887916e-05,
"loss": 0.3006,
"step": 6480
},
{
"epoch": 1.11,
"learning_rate": 1.7600262697022766e-05,
"loss": 0.3384,
"step": 6485
},
{
"epoch": 1.11,
"learning_rate": 1.7598073555166376e-05,
"loss": 0.2798,
"step": 6490
},
{
"epoch": 1.11,
"learning_rate": 1.7595884413309986e-05,
"loss": 0.1744,
"step": 6495
},
{
"epoch": 1.11,
"learning_rate": 1.7593695271453592e-05,
"loss": 0.0858,
"step": 6500
},
{
"epoch": 1.11,
"eval_loss": 0.09042953699827194,
"eval_runtime": 2254.4696,
"eval_samples_per_second": 9.204,
"eval_steps_per_second": 2.301,
"eval_wer": 0.3147622617262797,
"step": 6500
},
{
"epoch": 1.11,
"learning_rate": 1.75915061295972e-05,
"loss": 0.0949,
"step": 6505
},
{
"epoch": 1.12,
"learning_rate": 1.758931698774081e-05,
"loss": 0.1514,
"step": 6510
},
{
"epoch": 1.12,
"learning_rate": 1.7587127845884415e-05,
"loss": 0.1601,
"step": 6515
},
{
"epoch": 1.12,
"learning_rate": 1.758493870402802e-05,
"loss": 0.191,
"step": 6520
},
{
"epoch": 1.12,
"learning_rate": 1.758274956217163e-05,
"loss": 0.2281,
"step": 6525
},
{
"epoch": 1.12,
"learning_rate": 1.7580560420315238e-05,
"loss": 0.2859,
"step": 6530
},
{
"epoch": 1.12,
"learning_rate": 1.7578371278458844e-05,
"loss": 0.3878,
"step": 6535
},
{
"epoch": 1.12,
"learning_rate": 1.7576182136602454e-05,
"loss": 0.2493,
"step": 6540
},
{
"epoch": 1.12,
"learning_rate": 1.757399299474606e-05,
"loss": 0.2114,
"step": 6545
},
{
"epoch": 1.12,
"learning_rate": 1.7571803852889667e-05,
"loss": 0.0827,
"step": 6550
},
{
"epoch": 1.12,
"learning_rate": 1.7569614711033277e-05,
"loss": 0.1021,
"step": 6555
},
{
"epoch": 1.12,
"learning_rate": 1.7567425569176883e-05,
"loss": 0.1431,
"step": 6560
},
{
"epoch": 1.13,
"learning_rate": 1.7565236427320493e-05,
"loss": 0.1227,
"step": 6565
},
{
"epoch": 1.13,
"learning_rate": 1.75630472854641e-05,
"loss": 0.1769,
"step": 6570
},
{
"epoch": 1.13,
"learning_rate": 1.7560858143607706e-05,
"loss": 0.1854,
"step": 6575
},
{
"epoch": 1.13,
"learning_rate": 1.7558669001751315e-05,
"loss": 0.2238,
"step": 6580
},
{
"epoch": 1.13,
"learning_rate": 1.7556479859894922e-05,
"loss": 0.4255,
"step": 6585
},
{
"epoch": 1.13,
"learning_rate": 1.7554290718038532e-05,
"loss": 0.3124,
"step": 6590
},
{
"epoch": 1.13,
"learning_rate": 1.7552101576182138e-05,
"loss": 0.1814,
"step": 6595
},
{
"epoch": 1.13,
"learning_rate": 1.7549912434325748e-05,
"loss": 0.0873,
"step": 6600
},
{
"epoch": 1.13,
"learning_rate": 1.7547723292469354e-05,
"loss": 0.112,
"step": 6605
},
{
"epoch": 1.13,
"learning_rate": 1.754553415061296e-05,
"loss": 0.0944,
"step": 6610
},
{
"epoch": 1.13,
"learning_rate": 1.754334500875657e-05,
"loss": 0.1604,
"step": 6615
},
{
"epoch": 1.13,
"learning_rate": 1.7541155866900177e-05,
"loss": 0.1616,
"step": 6620
},
{
"epoch": 1.14,
"learning_rate": 1.7538966725043784e-05,
"loss": 0.189,
"step": 6625
},
{
"epoch": 1.14,
"learning_rate": 1.7536777583187393e-05,
"loss": 0.2033,
"step": 6630
},
{
"epoch": 1.14,
"learning_rate": 1.7534588441331e-05,
"loss": 0.3847,
"step": 6635
},
{
"epoch": 1.14,
"learning_rate": 1.7532399299474606e-05,
"loss": 0.2659,
"step": 6640
},
{
"epoch": 1.14,
"learning_rate": 1.7530210157618216e-05,
"loss": 0.1836,
"step": 6645
},
{
"epoch": 1.14,
"learning_rate": 1.7528021015761823e-05,
"loss": 0.0933,
"step": 6650
},
{
"epoch": 1.14,
"learning_rate": 1.752583187390543e-05,
"loss": 0.0976,
"step": 6655
},
{
"epoch": 1.14,
"learning_rate": 1.752364273204904e-05,
"loss": 0.1242,
"step": 6660
},
{
"epoch": 1.14,
"learning_rate": 1.7521453590192645e-05,
"loss": 0.1529,
"step": 6665
},
{
"epoch": 1.14,
"learning_rate": 1.751926444833625e-05,
"loss": 0.1521,
"step": 6670
},
{
"epoch": 1.14,
"learning_rate": 1.751707530647986e-05,
"loss": 0.1873,
"step": 6675
},
{
"epoch": 1.14,
"learning_rate": 1.7514886164623468e-05,
"loss": 0.1862,
"step": 6680
},
{
"epoch": 1.15,
"learning_rate": 1.7512697022767078e-05,
"loss": 0.3437,
"step": 6685
},
{
"epoch": 1.15,
"learning_rate": 1.7510507880910684e-05,
"loss": 0.2813,
"step": 6690
},
{
"epoch": 1.15,
"learning_rate": 1.7508318739054294e-05,
"loss": 0.166,
"step": 6695
},
{
"epoch": 1.15,
"learning_rate": 1.75061295971979e-05,
"loss": 0.091,
"step": 6700
},
{
"epoch": 1.15,
"learning_rate": 1.7503940455341507e-05,
"loss": 0.0927,
"step": 6705
},
{
"epoch": 1.15,
"learning_rate": 1.7501751313485117e-05,
"loss": 0.1468,
"step": 6710
},
{
"epoch": 1.15,
"learning_rate": 1.7499562171628723e-05,
"loss": 0.1188,
"step": 6715
},
{
"epoch": 1.15,
"learning_rate": 1.749737302977233e-05,
"loss": 0.1618,
"step": 6720
},
{
"epoch": 1.15,
"learning_rate": 1.749518388791594e-05,
"loss": 0.1921,
"step": 6725
},
{
"epoch": 1.15,
"learning_rate": 1.7492994746059546e-05,
"loss": 0.2274,
"step": 6730
},
{
"epoch": 1.15,
"learning_rate": 1.7490805604203152e-05,
"loss": 0.368,
"step": 6735
},
{
"epoch": 1.16,
"learning_rate": 1.7488616462346762e-05,
"loss": 0.2634,
"step": 6740
},
{
"epoch": 1.16,
"learning_rate": 1.748642732049037e-05,
"loss": 0.1524,
"step": 6745
},
{
"epoch": 1.16,
"learning_rate": 1.7484238178633975e-05,
"loss": 0.0747,
"step": 6750
},
{
"epoch": 1.16,
"learning_rate": 1.7482049036777585e-05,
"loss": 0.0798,
"step": 6755
},
{
"epoch": 1.16,
"learning_rate": 1.747985989492119e-05,
"loss": 0.1066,
"step": 6760
},
{
"epoch": 1.16,
"learning_rate": 1.7477670753064798e-05,
"loss": 0.1479,
"step": 6765
},
{
"epoch": 1.16,
"learning_rate": 1.7475481611208407e-05,
"loss": 0.1614,
"step": 6770
},
{
"epoch": 1.16,
"learning_rate": 1.7473292469352014e-05,
"loss": 0.2076,
"step": 6775
},
{
"epoch": 1.16,
"learning_rate": 1.7471103327495624e-05,
"loss": 0.2433,
"step": 6780
},
{
"epoch": 1.16,
"learning_rate": 1.7468914185639234e-05,
"loss": 0.4237,
"step": 6785
},
{
"epoch": 1.16,
"learning_rate": 1.746672504378284e-05,
"loss": 0.2147,
"step": 6790
},
{
"epoch": 1.16,
"learning_rate": 1.7464535901926446e-05,
"loss": 0.218,
"step": 6795
},
{
"epoch": 1.17,
"learning_rate": 1.7462346760070056e-05,
"loss": 0.0865,
"step": 6800
},
{
"epoch": 1.17,
"learning_rate": 1.7460157618213663e-05,
"loss": 0.0821,
"step": 6805
},
{
"epoch": 1.17,
"learning_rate": 1.745796847635727e-05,
"loss": 0.1496,
"step": 6810
},
{
"epoch": 1.17,
"learning_rate": 1.745577933450088e-05,
"loss": 0.1664,
"step": 6815
},
{
"epoch": 1.17,
"learning_rate": 1.7453590192644485e-05,
"loss": 0.149,
"step": 6820
},
{
"epoch": 1.17,
"learning_rate": 1.7451401050788092e-05,
"loss": 0.1445,
"step": 6825
},
{
"epoch": 1.17,
"learning_rate": 1.74492119089317e-05,
"loss": 0.1614,
"step": 6830
},
{
"epoch": 1.17,
"learning_rate": 1.7447022767075308e-05,
"loss": 0.424,
"step": 6835
},
{
"epoch": 1.17,
"learning_rate": 1.7444833625218915e-05,
"loss": 0.2584,
"step": 6840
},
{
"epoch": 1.17,
"learning_rate": 1.7442644483362524e-05,
"loss": 0.2146,
"step": 6845
},
{
"epoch": 1.17,
"learning_rate": 1.744045534150613e-05,
"loss": 0.0793,
"step": 6850
},
{
"epoch": 1.17,
"learning_rate": 1.7438266199649737e-05,
"loss": 0.1003,
"step": 6855
},
{
"epoch": 1.18,
"learning_rate": 1.7436077057793347e-05,
"loss": 0.1263,
"step": 6860
},
{
"epoch": 1.18,
"learning_rate": 1.7433887915936953e-05,
"loss": 0.1552,
"step": 6865
},
{
"epoch": 1.18,
"learning_rate": 1.743169877408056e-05,
"loss": 0.1546,
"step": 6870
},
{
"epoch": 1.18,
"learning_rate": 1.742950963222417e-05,
"loss": 0.1368,
"step": 6875
},
{
"epoch": 1.18,
"learning_rate": 1.7427320490367776e-05,
"loss": 0.247,
"step": 6880
},
{
"epoch": 1.18,
"learning_rate": 1.7425131348511386e-05,
"loss": 0.3711,
"step": 6885
},
{
"epoch": 1.18,
"learning_rate": 1.7422942206654992e-05,
"loss": 0.2742,
"step": 6890
},
{
"epoch": 1.18,
"learning_rate": 1.7420753064798602e-05,
"loss": 0.1673,
"step": 6895
},
{
"epoch": 1.18,
"learning_rate": 1.741856392294221e-05,
"loss": 0.0815,
"step": 6900
},
{
"epoch": 1.18,
"learning_rate": 1.7416374781085815e-05,
"loss": 0.0807,
"step": 6905
},
{
"epoch": 1.18,
"learning_rate": 1.7414185639229425e-05,
"loss": 0.1381,
"step": 6910
},
{
"epoch": 1.19,
"learning_rate": 1.741199649737303e-05,
"loss": 0.1571,
"step": 6915
},
{
"epoch": 1.19,
"learning_rate": 1.7409807355516638e-05,
"loss": 0.1556,
"step": 6920
},
{
"epoch": 1.19,
"learning_rate": 1.7407618213660248e-05,
"loss": 0.1736,
"step": 6925
},
{
"epoch": 1.19,
"learning_rate": 1.7405429071803854e-05,
"loss": 0.1894,
"step": 6930
},
{
"epoch": 1.19,
"learning_rate": 1.740323992994746e-05,
"loss": 0.3413,
"step": 6935
},
{
"epoch": 1.19,
"learning_rate": 1.740105078809107e-05,
"loss": 0.267,
"step": 6940
},
{
"epoch": 1.19,
"learning_rate": 1.7398861646234677e-05,
"loss": 0.1771,
"step": 6945
},
{
"epoch": 1.19,
"learning_rate": 1.7396672504378283e-05,
"loss": 0.0876,
"step": 6950
},
{
"epoch": 1.19,
"learning_rate": 1.7394483362521893e-05,
"loss": 0.1044,
"step": 6955
},
{
"epoch": 1.19,
"learning_rate": 1.73922942206655e-05,
"loss": 0.1168,
"step": 6960
},
{
"epoch": 1.19,
"learning_rate": 1.7390105078809106e-05,
"loss": 0.139,
"step": 6965
},
{
"epoch": 1.19,
"learning_rate": 1.7387915936952716e-05,
"loss": 0.1457,
"step": 6970
},
{
"epoch": 1.2,
"learning_rate": 1.7385726795096322e-05,
"loss": 0.2225,
"step": 6975
},
{
"epoch": 1.2,
"learning_rate": 1.7383537653239932e-05,
"loss": 0.2561,
"step": 6980
},
{
"epoch": 1.2,
"learning_rate": 1.738134851138354e-05,
"loss": 0.3332,
"step": 6985
},
{
"epoch": 1.2,
"learning_rate": 1.7379159369527148e-05,
"loss": 0.2488,
"step": 6990
},
{
"epoch": 1.2,
"learning_rate": 1.7376970227670755e-05,
"loss": 0.1535,
"step": 6995
},
{
"epoch": 1.2,
"learning_rate": 1.737478108581436e-05,
"loss": 0.0719,
"step": 7000
},
{
"epoch": 1.2,
"eval_loss": 0.062354039400815964,
"eval_runtime": 2227.896,
"eval_samples_per_second": 9.314,
"eval_steps_per_second": 2.329,
"eval_wer": 0.26716106232597986,
"step": 7000
},
{
"epoch": 1.2,
"learning_rate": 1.737259194395797e-05,
"loss": 0.1103,
"step": 7005
},
{
"epoch": 1.2,
"learning_rate": 1.7370402802101577e-05,
"loss": 0.1154,
"step": 7010
},
{
"epoch": 1.2,
"learning_rate": 1.7368213660245187e-05,
"loss": 0.1189,
"step": 7015
},
{
"epoch": 1.2,
"learning_rate": 1.7366024518388794e-05,
"loss": 0.2015,
"step": 7020
},
{
"epoch": 1.2,
"learning_rate": 1.73638353765324e-05,
"loss": 0.1962,
"step": 7025
},
{
"epoch": 1.2,
"learning_rate": 1.736164623467601e-05,
"loss": 0.2081,
"step": 7030
},
{
"epoch": 1.21,
"learning_rate": 1.7359457092819616e-05,
"loss": 0.3464,
"step": 7035
},
{
"epoch": 1.21,
"learning_rate": 1.7357267950963223e-05,
"loss": 0.2744,
"step": 7040
},
{
"epoch": 1.21,
"learning_rate": 1.7355078809106833e-05,
"loss": 0.1355,
"step": 7045
},
{
"epoch": 1.21,
"learning_rate": 1.735288966725044e-05,
"loss": 0.0653,
"step": 7050
},
{
"epoch": 1.21,
"learning_rate": 1.7350700525394046e-05,
"loss": 0.0964,
"step": 7055
},
{
"epoch": 1.21,
"learning_rate": 1.7348511383537655e-05,
"loss": 0.1144,
"step": 7060
},
{
"epoch": 1.21,
"learning_rate": 1.7346322241681262e-05,
"loss": 0.1422,
"step": 7065
},
{
"epoch": 1.21,
"learning_rate": 1.7344133099824868e-05,
"loss": 0.1485,
"step": 7070
},
{
"epoch": 1.21,
"learning_rate": 1.7341943957968478e-05,
"loss": 0.2437,
"step": 7075
},
{
"epoch": 1.21,
"learning_rate": 1.7339754816112088e-05,
"loss": 0.2249,
"step": 7080
},
{
"epoch": 1.21,
"learning_rate": 1.7337565674255694e-05,
"loss": 0.3723,
"step": 7085
},
{
"epoch": 1.22,
"learning_rate": 1.73353765323993e-05,
"loss": 0.2405,
"step": 7090
},
{
"epoch": 1.22,
"learning_rate": 1.733318739054291e-05,
"loss": 0.2116,
"step": 7095
},
{
"epoch": 1.22,
"learning_rate": 1.7330998248686517e-05,
"loss": 0.0728,
"step": 7100
},
{
"epoch": 1.22,
"learning_rate": 1.7328809106830123e-05,
"loss": 0.0656,
"step": 7105
},
{
"epoch": 1.22,
"learning_rate": 1.7326619964973733e-05,
"loss": 0.0961,
"step": 7110
},
{
"epoch": 1.22,
"learning_rate": 1.732443082311734e-05,
"loss": 0.1147,
"step": 7115
},
{
"epoch": 1.22,
"learning_rate": 1.7322241681260946e-05,
"loss": 0.1297,
"step": 7120
},
{
"epoch": 1.22,
"learning_rate": 1.7320052539404556e-05,
"loss": 0.1589,
"step": 7125
},
{
"epoch": 1.22,
"learning_rate": 1.7317863397548162e-05,
"loss": 0.197,
"step": 7130
},
{
"epoch": 1.22,
"learning_rate": 1.731567425569177e-05,
"loss": 0.3673,
"step": 7135
},
{
"epoch": 1.22,
"learning_rate": 1.731348511383538e-05,
"loss": 0.2773,
"step": 7140
},
{
"epoch": 1.22,
"learning_rate": 1.7311295971978985e-05,
"loss": 0.2129,
"step": 7145
},
{
"epoch": 1.23,
"learning_rate": 1.730910683012259e-05,
"loss": 0.0786,
"step": 7150
},
{
"epoch": 1.23,
"learning_rate": 1.73069176882662e-05,
"loss": 0.0764,
"step": 7155
},
{
"epoch": 1.23,
"learning_rate": 1.7304728546409808e-05,
"loss": 0.1007,
"step": 7160
},
{
"epoch": 1.23,
"learning_rate": 1.7302539404553414e-05,
"loss": 0.1079,
"step": 7165
},
{
"epoch": 1.23,
"learning_rate": 1.7300350262697024e-05,
"loss": 0.1327,
"step": 7170
},
{
"epoch": 1.23,
"learning_rate": 1.729816112084063e-05,
"loss": 0.1521,
"step": 7175
},
{
"epoch": 1.23,
"learning_rate": 1.729597197898424e-05,
"loss": 0.2474,
"step": 7180
},
{
"epoch": 1.23,
"learning_rate": 1.7293782837127847e-05,
"loss": 0.4158,
"step": 7185
},
{
"epoch": 1.23,
"learning_rate": 1.7291593695271457e-05,
"loss": 0.2632,
"step": 7190
},
{
"epoch": 1.23,
"learning_rate": 1.7289404553415063e-05,
"loss": 0.1471,
"step": 7195
},
{
"epoch": 1.23,
"learning_rate": 1.728721541155867e-05,
"loss": 0.0835,
"step": 7200
},
{
"epoch": 1.23,
"learning_rate": 1.728502626970228e-05,
"loss": 0.0835,
"step": 7205
},
{
"epoch": 1.24,
"learning_rate": 1.7282837127845886e-05,
"loss": 0.1021,
"step": 7210
},
{
"epoch": 1.24,
"learning_rate": 1.7280647985989492e-05,
"loss": 0.1188,
"step": 7215
},
{
"epoch": 1.24,
"learning_rate": 1.7278458844133102e-05,
"loss": 0.138,
"step": 7220
},
{
"epoch": 1.24,
"learning_rate": 1.727626970227671e-05,
"loss": 0.2125,
"step": 7225
},
{
"epoch": 1.24,
"learning_rate": 1.7274080560420318e-05,
"loss": 0.2581,
"step": 7230
},
{
"epoch": 1.24,
"learning_rate": 1.7271891418563925e-05,
"loss": 0.3356,
"step": 7235
},
{
"epoch": 1.24,
"learning_rate": 1.726970227670753e-05,
"loss": 0.2451,
"step": 7240
},
{
"epoch": 1.24,
"learning_rate": 1.726751313485114e-05,
"loss": 0.149,
"step": 7245
},
{
"epoch": 1.24,
"learning_rate": 1.7265323992994747e-05,
"loss": 0.0895,
"step": 7250
},
{
"epoch": 1.24,
"learning_rate": 1.7263134851138354e-05,
"loss": 0.0878,
"step": 7255
},
{
"epoch": 1.24,
"learning_rate": 1.7260945709281964e-05,
"loss": 0.1021,
"step": 7260
},
{
"epoch": 1.25,
"learning_rate": 1.725875656742557e-05,
"loss": 0.1291,
"step": 7265
},
{
"epoch": 1.25,
"learning_rate": 1.7256567425569176e-05,
"loss": 0.1829,
"step": 7270
},
{
"epoch": 1.25,
"learning_rate": 1.7254378283712786e-05,
"loss": 0.2169,
"step": 7275
},
{
"epoch": 1.25,
"learning_rate": 1.7252189141856396e-05,
"loss": 0.2572,
"step": 7280
},
{
"epoch": 1.25,
"learning_rate": 1.7250000000000003e-05,
"loss": 0.3805,
"step": 7285
},
{
"epoch": 1.25,
"learning_rate": 1.724781085814361e-05,
"loss": 0.2442,
"step": 7290
},
{
"epoch": 1.25,
"learning_rate": 1.724562171628722e-05,
"loss": 0.1712,
"step": 7295
},
{
"epoch": 1.25,
"learning_rate": 1.7243432574430825e-05,
"loss": 0.0733,
"step": 7300
},
{
"epoch": 1.25,
"learning_rate": 1.724124343257443e-05,
"loss": 0.0892,
"step": 7305
},
{
"epoch": 1.25,
"learning_rate": 1.723905429071804e-05,
"loss": 0.0957,
"step": 7310
},
{
"epoch": 1.25,
"learning_rate": 1.7236865148861648e-05,
"loss": 0.1364,
"step": 7315
},
{
"epoch": 1.25,
"learning_rate": 1.7234676007005254e-05,
"loss": 0.138,
"step": 7320
},
{
"epoch": 1.26,
"learning_rate": 1.7232486865148864e-05,
"loss": 0.1247,
"step": 7325
},
{
"epoch": 1.26,
"learning_rate": 1.723029772329247e-05,
"loss": 0.2007,
"step": 7330
},
{
"epoch": 1.26,
"learning_rate": 1.7228108581436077e-05,
"loss": 0.3482,
"step": 7335
},
{
"epoch": 1.26,
"learning_rate": 1.7225919439579687e-05,
"loss": 0.2663,
"step": 7340
},
{
"epoch": 1.26,
"learning_rate": 1.7223730297723293e-05,
"loss": 0.204,
"step": 7345
},
{
"epoch": 1.26,
"learning_rate": 1.72215411558669e-05,
"loss": 0.0612,
"step": 7350
},
{
"epoch": 1.26,
"learning_rate": 1.721935201401051e-05,
"loss": 0.0848,
"step": 7355
},
{
"epoch": 1.26,
"learning_rate": 1.7217162872154116e-05,
"loss": 0.1046,
"step": 7360
},
{
"epoch": 1.26,
"learning_rate": 1.7214973730297722e-05,
"loss": 0.1505,
"step": 7365
},
{
"epoch": 1.26,
"learning_rate": 1.7212784588441332e-05,
"loss": 0.1623,
"step": 7370
},
{
"epoch": 1.26,
"learning_rate": 1.7210595446584942e-05,
"loss": 0.1867,
"step": 7375
},
{
"epoch": 1.26,
"learning_rate": 1.720840630472855e-05,
"loss": 0.2044,
"step": 7380
},
{
"epoch": 1.27,
"learning_rate": 1.7206217162872155e-05,
"loss": 0.4967,
"step": 7385
},
{
"epoch": 1.27,
"learning_rate": 1.7204028021015765e-05,
"loss": 0.2562,
"step": 7390
},
{
"epoch": 1.27,
"learning_rate": 1.720183887915937e-05,
"loss": 0.1785,
"step": 7395
},
{
"epoch": 1.27,
"learning_rate": 1.7199649737302978e-05,
"loss": 0.0839,
"step": 7400
},
{
"epoch": 1.27,
"learning_rate": 1.7197460595446588e-05,
"loss": 0.1055,
"step": 7405
},
{
"epoch": 1.27,
"learning_rate": 1.7195271453590194e-05,
"loss": 0.1497,
"step": 7410
},
{
"epoch": 1.27,
"learning_rate": 1.71930823117338e-05,
"loss": 0.1391,
"step": 7415
},
{
"epoch": 1.27,
"learning_rate": 1.719089316987741e-05,
"loss": 0.1839,
"step": 7420
},
{
"epoch": 1.27,
"learning_rate": 1.7188704028021017e-05,
"loss": 0.1571,
"step": 7425
},
{
"epoch": 1.27,
"learning_rate": 1.7186514886164623e-05,
"loss": 0.2292,
"step": 7430
},
{
"epoch": 1.27,
"learning_rate": 1.7184325744308233e-05,
"loss": 0.3703,
"step": 7435
},
{
"epoch": 1.28,
"learning_rate": 1.718213660245184e-05,
"loss": 0.2312,
"step": 7440
},
{
"epoch": 1.28,
"learning_rate": 1.717994746059545e-05,
"loss": 0.1651,
"step": 7445
},
{
"epoch": 1.28,
"learning_rate": 1.7177758318739056e-05,
"loss": 0.0865,
"step": 7450
},
{
"epoch": 1.28,
"learning_rate": 1.7175569176882662e-05,
"loss": 0.104,
"step": 7455
},
{
"epoch": 1.28,
"learning_rate": 1.7173380035026272e-05,
"loss": 0.1145,
"step": 7460
},
{
"epoch": 1.28,
"learning_rate": 1.7171190893169878e-05,
"loss": 0.1169,
"step": 7465
},
{
"epoch": 1.28,
"learning_rate": 1.7169001751313485e-05,
"loss": 0.1516,
"step": 7470
},
{
"epoch": 1.28,
"learning_rate": 1.7166812609457095e-05,
"loss": 0.1637,
"step": 7475
},
{
"epoch": 1.28,
"learning_rate": 1.7164623467600704e-05,
"loss": 0.2032,
"step": 7480
},
{
"epoch": 1.28,
"learning_rate": 1.716243432574431e-05,
"loss": 0.3624,
"step": 7485
},
{
"epoch": 1.28,
"learning_rate": 1.7160245183887917e-05,
"loss": 0.2835,
"step": 7490
},
{
"epoch": 1.28,
"learning_rate": 1.7158056042031527e-05,
"loss": 0.1469,
"step": 7495
},
{
"epoch": 1.29,
"learning_rate": 1.7155866900175134e-05,
"loss": 0.0694,
"step": 7500
},
{
"epoch": 1.29,
"eval_loss": 0.054694876074790955,
"eval_runtime": 1737.5048,
"eval_samples_per_second": 11.942,
"eval_steps_per_second": 2.986,
"eval_wer": 0.3052313129149711,
"step": 7500
}
],
"max_steps": 46680,
"num_train_epochs": 8,
"total_flos": 4.644733670770764e+18,
"trial_name": null,
"trial_params": null
}