whisper-lg-el-intlv-xs / trainer_state.json
farsipal's picture
Upload 22 files
c35582b
{
"best_metric": 9.899702823179792,
"best_model_checkpoint": "./data/finetuningRuns/whisper-lg-el-intlv-xs/checkpoint-4000",
"epoch": 24.875621890547265,
"global_step": 10000,
"is_hyper_param_search": false,
"is_local_process_zero": true,
"is_world_process_zero": true,
"log_history": [
{
"epoch": 0.06,
"learning_rate": 1.61e-07,
"loss": 0.5163,
"step": 25
},
{
"epoch": 0.12,
"learning_rate": 3.36e-07,
"loss": 0.4817,
"step": 50
},
{
"epoch": 0.19,
"learning_rate": 5.11e-07,
"loss": 0.3656,
"step": 75
},
{
"epoch": 0.25,
"learning_rate": 6.86e-07,
"loss": 0.2421,
"step": 100
},
{
"epoch": 0.31,
"learning_rate": 8.61e-07,
"loss": 0.2035,
"step": 125
},
{
"epoch": 0.37,
"learning_rate": 1.036e-06,
"loss": 0.1765,
"step": 150
},
{
"epoch": 0.44,
"learning_rate": 1.211e-06,
"loss": 0.1664,
"step": 175
},
{
"epoch": 0.5,
"learning_rate": 1.379e-06,
"loss": 0.1451,
"step": 200
},
{
"epoch": 0.56,
"learning_rate": 1.554e-06,
"loss": 0.1496,
"step": 225
},
{
"epoch": 0.62,
"learning_rate": 1.7289999999999999e-06,
"loss": 0.1385,
"step": 250
},
{
"epoch": 0.68,
"learning_rate": 1.9040000000000001e-06,
"loss": 0.1285,
"step": 275
},
{
"epoch": 0.75,
"learning_rate": 2.079e-06,
"loss": 0.1227,
"step": 300
},
{
"epoch": 0.81,
"learning_rate": 2.254e-06,
"loss": 0.1184,
"step": 325
},
{
"epoch": 0.87,
"learning_rate": 2.429e-06,
"loss": 0.113,
"step": 350
},
{
"epoch": 0.93,
"learning_rate": 2.604e-06,
"loss": 0.1087,
"step": 375
},
{
"epoch": 1.0,
"learning_rate": 2.779e-06,
"loss": 0.1128,
"step": 400
},
{
"epoch": 1.06,
"learning_rate": 2.954e-06,
"loss": 0.0815,
"step": 425
},
{
"epoch": 1.12,
"learning_rate": 3.129e-06,
"loss": 0.0883,
"step": 450
},
{
"epoch": 1.18,
"learning_rate": 3.3039999999999996e-06,
"loss": 0.0762,
"step": 475
},
{
"epoch": 1.24,
"learning_rate": 3.479e-06,
"loss": 0.0804,
"step": 500
},
{
"epoch": 1.31,
"learning_rate": 3.4918947368421053e-06,
"loss": 0.0724,
"step": 525
},
{
"epoch": 1.37,
"learning_rate": 3.482684210526316e-06,
"loss": 0.0726,
"step": 550
},
{
"epoch": 1.43,
"learning_rate": 3.4734736842105263e-06,
"loss": 0.0759,
"step": 575
},
{
"epoch": 1.49,
"learning_rate": 3.464263157894737e-06,
"loss": 0.0809,
"step": 600
},
{
"epoch": 1.55,
"learning_rate": 3.4550526315789474e-06,
"loss": 0.0745,
"step": 625
},
{
"epoch": 1.62,
"learning_rate": 3.4458421052631577e-06,
"loss": 0.0671,
"step": 650
},
{
"epoch": 1.68,
"learning_rate": 3.4366315789473685e-06,
"loss": 0.0679,
"step": 675
},
{
"epoch": 1.74,
"learning_rate": 3.427421052631579e-06,
"loss": 0.0728,
"step": 700
},
{
"epoch": 1.8,
"learning_rate": 3.4182105263157896e-06,
"loss": 0.061,
"step": 725
},
{
"epoch": 1.87,
"learning_rate": 3.409e-06,
"loss": 0.0619,
"step": 750
},
{
"epoch": 1.93,
"learning_rate": 3.39978947368421e-06,
"loss": 0.0672,
"step": 775
},
{
"epoch": 1.99,
"learning_rate": 3.3909473684210528e-06,
"loss": 0.0611,
"step": 800
},
{
"epoch": 2.05,
"learning_rate": 3.381736842105263e-06,
"loss": 0.0409,
"step": 825
},
{
"epoch": 2.11,
"learning_rate": 3.372526315789474e-06,
"loss": 0.032,
"step": 850
},
{
"epoch": 2.18,
"learning_rate": 3.363315789473684e-06,
"loss": 0.0308,
"step": 875
},
{
"epoch": 2.24,
"learning_rate": 3.354105263157895e-06,
"loss": 0.0318,
"step": 900
},
{
"epoch": 2.3,
"learning_rate": 3.3448947368421052e-06,
"loss": 0.035,
"step": 925
},
{
"epoch": 2.36,
"learning_rate": 3.3356842105263156e-06,
"loss": 0.0353,
"step": 950
},
{
"epoch": 2.43,
"learning_rate": 3.3264736842105263e-06,
"loss": 0.0301,
"step": 975
},
{
"epoch": 2.49,
"learning_rate": 3.3172631578947366e-06,
"loss": 0.0311,
"step": 1000
},
{
"epoch": 2.49,
"eval_loss": 0.1808706372976303,
"eval_runtime": 1982.2643,
"eval_samples_per_second": 0.856,
"eval_steps_per_second": 0.107,
"eval_wer": 10.549777117384844,
"step": 1000
},
{
"epoch": 2.55,
"learning_rate": 3.3080526315789474e-06,
"loss": 0.0309,
"step": 1025
},
{
"epoch": 2.61,
"learning_rate": 3.2988421052631577e-06,
"loss": 0.0364,
"step": 1050
},
{
"epoch": 2.67,
"learning_rate": 3.2896315789473685e-06,
"loss": 0.0323,
"step": 1075
},
{
"epoch": 2.74,
"learning_rate": 3.2804210526315788e-06,
"loss": 0.0322,
"step": 1100
},
{
"epoch": 2.8,
"learning_rate": 3.271210526315789e-06,
"loss": 0.0304,
"step": 1125
},
{
"epoch": 2.86,
"learning_rate": 3.2620000000000003e-06,
"loss": 0.0299,
"step": 1150
},
{
"epoch": 2.92,
"learning_rate": 3.2527894736842106e-06,
"loss": 0.0272,
"step": 1175
},
{
"epoch": 2.99,
"learning_rate": 3.243578947368421e-06,
"loss": 0.0241,
"step": 1200
},
{
"epoch": 3.05,
"learning_rate": 3.2343684210526317e-06,
"loss": 0.0186,
"step": 1225
},
{
"epoch": 3.11,
"learning_rate": 3.225157894736842e-06,
"loss": 0.015,
"step": 1250
},
{
"epoch": 3.17,
"learning_rate": 3.2159473684210527e-06,
"loss": 0.0125,
"step": 1275
},
{
"epoch": 3.23,
"learning_rate": 3.206736842105263e-06,
"loss": 0.0127,
"step": 1300
},
{
"epoch": 3.3,
"learning_rate": 3.197526315789474e-06,
"loss": 0.0144,
"step": 1325
},
{
"epoch": 3.36,
"learning_rate": 3.188315789473684e-06,
"loss": 0.0148,
"step": 1350
},
{
"epoch": 3.42,
"learning_rate": 3.1791052631578945e-06,
"loss": 0.0122,
"step": 1375
},
{
"epoch": 3.48,
"learning_rate": 3.1698947368421052e-06,
"loss": 0.0124,
"step": 1400
},
{
"epoch": 3.54,
"learning_rate": 3.1606842105263155e-06,
"loss": 0.0165,
"step": 1425
},
{
"epoch": 3.61,
"learning_rate": 3.151842105263158e-06,
"loss": 0.014,
"step": 1450
},
{
"epoch": 3.67,
"learning_rate": 3.1426315789473684e-06,
"loss": 0.0138,
"step": 1475
},
{
"epoch": 3.73,
"learning_rate": 3.133421052631579e-06,
"loss": 0.0142,
"step": 1500
},
{
"epoch": 3.79,
"learning_rate": 3.1242105263157895e-06,
"loss": 0.0136,
"step": 1525
},
{
"epoch": 3.86,
"learning_rate": 3.115e-06,
"loss": 0.0123,
"step": 1550
},
{
"epoch": 3.92,
"learning_rate": 3.1057894736842106e-06,
"loss": 0.0135,
"step": 1575
},
{
"epoch": 3.98,
"learning_rate": 3.096578947368421e-06,
"loss": 0.0135,
"step": 1600
},
{
"epoch": 4.04,
"learning_rate": 3.0873684210526316e-06,
"loss": 0.0102,
"step": 1625
},
{
"epoch": 4.1,
"learning_rate": 3.078157894736842e-06,
"loss": 0.0061,
"step": 1650
},
{
"epoch": 4.17,
"learning_rate": 3.0689473684210523e-06,
"loss": 0.0064,
"step": 1675
},
{
"epoch": 4.23,
"learning_rate": 3.059736842105263e-06,
"loss": 0.0068,
"step": 1700
},
{
"epoch": 4.29,
"learning_rate": 3.0505263157894734e-06,
"loss": 0.0056,
"step": 1725
},
{
"epoch": 4.35,
"learning_rate": 3.041315789473684e-06,
"loss": 0.0062,
"step": 1750
},
{
"epoch": 4.42,
"learning_rate": 3.032105263157895e-06,
"loss": 0.0071,
"step": 1775
},
{
"epoch": 4.48,
"learning_rate": 3.022894736842105e-06,
"loss": 0.0061,
"step": 1800
},
{
"epoch": 4.54,
"learning_rate": 3.013684210526316e-06,
"loss": 0.0069,
"step": 1825
},
{
"epoch": 4.6,
"learning_rate": 3.0044736842105263e-06,
"loss": 0.0067,
"step": 1850
},
{
"epoch": 4.66,
"learning_rate": 2.995263157894737e-06,
"loss": 0.0055,
"step": 1875
},
{
"epoch": 4.73,
"learning_rate": 2.9860526315789473e-06,
"loss": 0.0067,
"step": 1900
},
{
"epoch": 4.79,
"learning_rate": 2.9768421052631577e-06,
"loss": 0.0061,
"step": 1925
},
{
"epoch": 4.85,
"learning_rate": 2.9676315789473684e-06,
"loss": 0.0074,
"step": 1950
},
{
"epoch": 4.91,
"learning_rate": 2.9584210526315787e-06,
"loss": 0.0087,
"step": 1975
},
{
"epoch": 4.98,
"learning_rate": 2.9492105263157895e-06,
"loss": 0.0074,
"step": 2000
},
{
"epoch": 4.98,
"eval_loss": 0.2470209300518036,
"eval_runtime": 1972.083,
"eval_samples_per_second": 0.86,
"eval_steps_per_second": 0.108,
"eval_wer": 10.280460624071322,
"step": 2000
},
{
"epoch": 5.04,
"learning_rate": 2.94e-06,
"loss": 0.0049,
"step": 2025
},
{
"epoch": 5.1,
"learning_rate": 2.9307894736842106e-06,
"loss": 0.0035,
"step": 2050
},
{
"epoch": 5.16,
"learning_rate": 2.921578947368421e-06,
"loss": 0.0045,
"step": 2075
},
{
"epoch": 5.22,
"learning_rate": 2.9123684210526316e-06,
"loss": 0.0038,
"step": 2100
},
{
"epoch": 5.29,
"learning_rate": 2.9031578947368424e-06,
"loss": 0.0033,
"step": 2125
},
{
"epoch": 5.35,
"learning_rate": 2.8939473684210527e-06,
"loss": 0.0036,
"step": 2150
},
{
"epoch": 5.41,
"learning_rate": 2.884736842105263e-06,
"loss": 0.0039,
"step": 2175
},
{
"epoch": 5.47,
"learning_rate": 2.8755263157894738e-06,
"loss": 0.0035,
"step": 2200
},
{
"epoch": 5.53,
"learning_rate": 2.866315789473684e-06,
"loss": 0.0038,
"step": 2225
},
{
"epoch": 5.6,
"learning_rate": 2.857105263157895e-06,
"loss": 0.0039,
"step": 2250
},
{
"epoch": 5.66,
"learning_rate": 2.847894736842105e-06,
"loss": 0.0036,
"step": 2275
},
{
"epoch": 5.72,
"learning_rate": 2.838684210526316e-06,
"loss": 0.0043,
"step": 2300
},
{
"epoch": 5.78,
"learning_rate": 2.8294736842105262e-06,
"loss": 0.0049,
"step": 2325
},
{
"epoch": 5.85,
"learning_rate": 2.8202631578947366e-06,
"loss": 0.0041,
"step": 2350
},
{
"epoch": 5.91,
"learning_rate": 2.8110526315789473e-06,
"loss": 0.004,
"step": 2375
},
{
"epoch": 5.97,
"learning_rate": 2.8018421052631576e-06,
"loss": 0.0036,
"step": 2400
},
{
"epoch": 6.03,
"learning_rate": 2.7926315789473684e-06,
"loss": 0.003,
"step": 2425
},
{
"epoch": 6.09,
"learning_rate": 2.783421052631579e-06,
"loss": 0.0026,
"step": 2450
},
{
"epoch": 6.16,
"learning_rate": 2.7742105263157895e-06,
"loss": 0.0019,
"step": 2475
},
{
"epoch": 6.22,
"learning_rate": 2.765e-06,
"loss": 0.0023,
"step": 2500
},
{
"epoch": 6.28,
"learning_rate": 2.7557894736842105e-06,
"loss": 0.0029,
"step": 2525
},
{
"epoch": 6.34,
"learning_rate": 2.7465789473684213e-06,
"loss": 0.0032,
"step": 2550
},
{
"epoch": 6.41,
"learning_rate": 2.7373684210526316e-06,
"loss": 0.0028,
"step": 2575
},
{
"epoch": 6.47,
"learning_rate": 2.728157894736842e-06,
"loss": 0.0025,
"step": 2600
},
{
"epoch": 6.53,
"learning_rate": 2.7189473684210527e-06,
"loss": 0.0028,
"step": 2625
},
{
"epoch": 6.59,
"learning_rate": 2.709736842105263e-06,
"loss": 0.0025,
"step": 2650
},
{
"epoch": 6.65,
"learning_rate": 2.7005263157894738e-06,
"loss": 0.0024,
"step": 2675
},
{
"epoch": 6.72,
"learning_rate": 2.691315789473684e-06,
"loss": 0.0032,
"step": 2700
},
{
"epoch": 6.78,
"learning_rate": 2.6821052631578944e-06,
"loss": 0.0031,
"step": 2725
},
{
"epoch": 6.84,
"learning_rate": 2.672894736842105e-06,
"loss": 0.0032,
"step": 2750
},
{
"epoch": 6.9,
"learning_rate": 2.6636842105263155e-06,
"loss": 0.0034,
"step": 2775
},
{
"epoch": 6.97,
"learning_rate": 2.6544736842105266e-06,
"loss": 0.0026,
"step": 2800
},
{
"epoch": 7.03,
"learning_rate": 2.645263157894737e-06,
"loss": 0.003,
"step": 2825
},
{
"epoch": 7.09,
"learning_rate": 2.6360526315789473e-06,
"loss": 0.003,
"step": 2850
},
{
"epoch": 7.15,
"learning_rate": 2.626842105263158e-06,
"loss": 0.0023,
"step": 2875
},
{
"epoch": 7.21,
"learning_rate": 2.6176315789473684e-06,
"loss": 0.0024,
"step": 2900
},
{
"epoch": 7.28,
"learning_rate": 2.608421052631579e-06,
"loss": 0.0019,
"step": 2925
},
{
"epoch": 7.34,
"learning_rate": 2.5992105263157894e-06,
"loss": 0.0021,
"step": 2950
},
{
"epoch": 7.4,
"learning_rate": 2.5899999999999998e-06,
"loss": 0.0019,
"step": 2975
},
{
"epoch": 7.46,
"learning_rate": 2.5807894736842105e-06,
"loss": 0.0019,
"step": 3000
},
{
"epoch": 7.46,
"eval_loss": 0.30083394050598145,
"eval_runtime": 1980.8887,
"eval_samples_per_second": 0.856,
"eval_steps_per_second": 0.107,
"eval_wer": 10.029717682020802,
"step": 3000
},
{
"epoch": 7.52,
"learning_rate": 2.571578947368421e-06,
"loss": 0.0022,
"step": 3025
},
{
"epoch": 7.59,
"learning_rate": 2.5623684210526316e-06,
"loss": 0.0022,
"step": 3050
},
{
"epoch": 7.65,
"learning_rate": 2.553157894736842e-06,
"loss": 0.0021,
"step": 3075
},
{
"epoch": 7.71,
"learning_rate": 2.5439473684210522e-06,
"loss": 0.0027,
"step": 3100
},
{
"epoch": 7.77,
"learning_rate": 2.5347368421052634e-06,
"loss": 0.0023,
"step": 3125
},
{
"epoch": 7.84,
"learning_rate": 2.5255263157894737e-06,
"loss": 0.0026,
"step": 3150
},
{
"epoch": 7.9,
"learning_rate": 2.5163157894736845e-06,
"loss": 0.0022,
"step": 3175
},
{
"epoch": 7.96,
"learning_rate": 2.507105263157895e-06,
"loss": 0.0023,
"step": 3200
},
{
"epoch": 8.02,
"learning_rate": 2.497894736842105e-06,
"loss": 0.0023,
"step": 3225
},
{
"epoch": 8.08,
"learning_rate": 2.488684210526316e-06,
"loss": 0.0018,
"step": 3250
},
{
"epoch": 8.15,
"learning_rate": 2.479473684210526e-06,
"loss": 0.0015,
"step": 3275
},
{
"epoch": 8.21,
"learning_rate": 2.470263157894737e-06,
"loss": 0.0016,
"step": 3300
},
{
"epoch": 8.27,
"learning_rate": 2.4610526315789473e-06,
"loss": 0.0016,
"step": 3325
},
{
"epoch": 8.33,
"learning_rate": 2.4518421052631576e-06,
"loss": 0.0016,
"step": 3350
},
{
"epoch": 8.4,
"learning_rate": 2.4426315789473684e-06,
"loss": 0.0014,
"step": 3375
},
{
"epoch": 8.46,
"learning_rate": 2.4334210526315787e-06,
"loss": 0.0015,
"step": 3400
},
{
"epoch": 8.52,
"learning_rate": 2.4242105263157894e-06,
"loss": 0.0016,
"step": 3425
},
{
"epoch": 8.58,
"learning_rate": 2.4149999999999997e-06,
"loss": 0.0019,
"step": 3450
},
{
"epoch": 8.64,
"learning_rate": 2.4057894736842105e-06,
"loss": 0.0015,
"step": 3475
},
{
"epoch": 8.71,
"learning_rate": 2.3965789473684212e-06,
"loss": 0.0022,
"step": 3500
},
{
"epoch": 8.77,
"learning_rate": 2.3873684210526316e-06,
"loss": 0.0018,
"step": 3525
},
{
"epoch": 8.83,
"learning_rate": 2.3781578947368423e-06,
"loss": 0.0027,
"step": 3550
},
{
"epoch": 8.89,
"learning_rate": 2.3689473684210526e-06,
"loss": 0.0016,
"step": 3575
},
{
"epoch": 8.96,
"learning_rate": 2.359736842105263e-06,
"loss": 0.0016,
"step": 3600
},
{
"epoch": 9.02,
"learning_rate": 2.3505263157894737e-06,
"loss": 0.0015,
"step": 3625
},
{
"epoch": 9.08,
"learning_rate": 2.341315789473684e-06,
"loss": 0.001,
"step": 3650
},
{
"epoch": 9.14,
"learning_rate": 2.3321052631578948e-06,
"loss": 0.0011,
"step": 3675
},
{
"epoch": 9.2,
"learning_rate": 2.322894736842105e-06,
"loss": 0.0011,
"step": 3700
},
{
"epoch": 9.27,
"learning_rate": 2.3136842105263154e-06,
"loss": 0.0012,
"step": 3725
},
{
"epoch": 9.33,
"learning_rate": 2.304473684210526e-06,
"loss": 0.0012,
"step": 3750
},
{
"epoch": 9.39,
"learning_rate": 2.2952631578947365e-06,
"loss": 0.0012,
"step": 3775
},
{
"epoch": 9.45,
"learning_rate": 2.2860526315789477e-06,
"loss": 0.0011,
"step": 3800
},
{
"epoch": 9.51,
"learning_rate": 2.276842105263158e-06,
"loss": 0.0009,
"step": 3825
},
{
"epoch": 9.58,
"learning_rate": 2.2676315789473683e-06,
"loss": 0.0009,
"step": 3850
},
{
"epoch": 9.64,
"learning_rate": 2.258421052631579e-06,
"loss": 0.0014,
"step": 3875
},
{
"epoch": 9.7,
"learning_rate": 2.2492105263157894e-06,
"loss": 0.0008,
"step": 3900
},
{
"epoch": 9.76,
"learning_rate": 2.24e-06,
"loss": 0.0015,
"step": 3925
},
{
"epoch": 9.83,
"learning_rate": 2.2307894736842105e-06,
"loss": 0.0009,
"step": 3950
},
{
"epoch": 9.89,
"learning_rate": 2.221578947368421e-06,
"loss": 0.0012,
"step": 3975
},
{
"epoch": 9.95,
"learning_rate": 2.2123684210526316e-06,
"loss": 0.0011,
"step": 4000
},
{
"epoch": 9.95,
"eval_loss": 0.29131078720092773,
"eval_runtime": 1980.7982,
"eval_samples_per_second": 0.856,
"eval_steps_per_second": 0.107,
"eval_wer": 9.899702823179792,
"step": 4000
},
{
"epoch": 10.01,
"learning_rate": 2.203157894736842e-06,
"loss": 0.0009,
"step": 4025
},
{
"epoch": 10.07,
"learning_rate": 2.1939473684210526e-06,
"loss": 0.0007,
"step": 4050
},
{
"epoch": 10.14,
"learning_rate": 2.184736842105263e-06,
"loss": 0.0007,
"step": 4075
},
{
"epoch": 10.2,
"learning_rate": 2.1755263157894737e-06,
"loss": 0.0015,
"step": 4100
},
{
"epoch": 10.26,
"learning_rate": 2.166315789473684e-06,
"loss": 0.0014,
"step": 4125
},
{
"epoch": 10.32,
"learning_rate": 2.1571052631578948e-06,
"loss": 0.001,
"step": 4150
},
{
"epoch": 10.39,
"learning_rate": 2.1478947368421055e-06,
"loss": 0.0013,
"step": 4175
},
{
"epoch": 10.45,
"learning_rate": 2.138684210526316e-06,
"loss": 0.0012,
"step": 4200
},
{
"epoch": 10.51,
"learning_rate": 2.129473684210526e-06,
"loss": 0.0008,
"step": 4225
},
{
"epoch": 10.57,
"learning_rate": 2.120263157894737e-06,
"loss": 0.0009,
"step": 4250
},
{
"epoch": 10.63,
"learning_rate": 2.1110526315789472e-06,
"loss": 0.0013,
"step": 4275
},
{
"epoch": 10.7,
"learning_rate": 2.101842105263158e-06,
"loss": 0.0012,
"step": 4300
},
{
"epoch": 10.76,
"learning_rate": 2.0926315789473683e-06,
"loss": 0.0007,
"step": 4325
},
{
"epoch": 10.82,
"learning_rate": 2.083421052631579e-06,
"loss": 0.001,
"step": 4350
},
{
"epoch": 10.88,
"learning_rate": 2.0742105263157894e-06,
"loss": 0.0007,
"step": 4375
},
{
"epoch": 10.95,
"learning_rate": 2.0649999999999997e-06,
"loss": 0.0007,
"step": 4400
},
{
"epoch": 11.01,
"learning_rate": 2.0557894736842105e-06,
"loss": 0.0006,
"step": 4425
},
{
"epoch": 11.07,
"learning_rate": 2.0465789473684208e-06,
"loss": 0.0005,
"step": 4450
},
{
"epoch": 11.13,
"learning_rate": 2.0373684210526315e-06,
"loss": 0.0005,
"step": 4475
},
{
"epoch": 11.19,
"learning_rate": 2.0281578947368423e-06,
"loss": 0.0005,
"step": 4500
},
{
"epoch": 11.26,
"learning_rate": 2.0189473684210526e-06,
"loss": 0.0008,
"step": 4525
},
{
"epoch": 11.32,
"learning_rate": 2.0097368421052634e-06,
"loss": 0.0009,
"step": 4550
},
{
"epoch": 11.38,
"learning_rate": 2.0005263157894737e-06,
"loss": 0.0009,
"step": 4575
},
{
"epoch": 11.44,
"learning_rate": 1.9913157894736844e-06,
"loss": 0.0009,
"step": 4600
},
{
"epoch": 11.5,
"learning_rate": 1.9821052631578948e-06,
"loss": 0.0007,
"step": 4625
},
{
"epoch": 11.57,
"learning_rate": 1.972894736842105e-06,
"loss": 0.0005,
"step": 4650
},
{
"epoch": 11.63,
"learning_rate": 1.963684210526316e-06,
"loss": 0.0005,
"step": 4675
},
{
"epoch": 11.69,
"learning_rate": 1.954473684210526e-06,
"loss": 0.0014,
"step": 4700
},
{
"epoch": 11.75,
"learning_rate": 1.945263157894737e-06,
"loss": 0.0011,
"step": 4725
},
{
"epoch": 11.82,
"learning_rate": 1.9360526315789472e-06,
"loss": 0.0007,
"step": 4750
},
{
"epoch": 11.88,
"learning_rate": 1.9268421052631575e-06,
"loss": 0.0019,
"step": 4775
},
{
"epoch": 11.94,
"learning_rate": 1.9176315789473683e-06,
"loss": 0.0008,
"step": 4800
},
{
"epoch": 12.0,
"learning_rate": 1.908421052631579e-06,
"loss": 0.0009,
"step": 4825
},
{
"epoch": 12.06,
"learning_rate": 1.8992105263157896e-06,
"loss": 0.0017,
"step": 4850
},
{
"epoch": 12.13,
"learning_rate": 1.8900000000000001e-06,
"loss": 0.001,
"step": 4875
},
{
"epoch": 12.19,
"learning_rate": 1.8807894736842107e-06,
"loss": 0.0009,
"step": 4900
},
{
"epoch": 12.25,
"learning_rate": 1.8715789473684212e-06,
"loss": 0.0008,
"step": 4925
},
{
"epoch": 12.31,
"learning_rate": 1.8623684210526315e-06,
"loss": 0.001,
"step": 4950
},
{
"epoch": 12.38,
"learning_rate": 1.853157894736842e-06,
"loss": 0.0009,
"step": 4975
},
{
"epoch": 12.44,
"learning_rate": 1.8439473684210526e-06,
"loss": 0.0009,
"step": 5000
},
{
"epoch": 12.44,
"eval_loss": 0.3091869652271271,
"eval_runtime": 1985.8241,
"eval_samples_per_second": 0.854,
"eval_steps_per_second": 0.107,
"eval_wer": 10.187592867756315,
"step": 5000
},
{
"epoch": 12.5,
"learning_rate": 1.8347368421052631e-06,
"loss": 0.0006,
"step": 5025
},
{
"epoch": 12.56,
"learning_rate": 1.8255263157894737e-06,
"loss": 0.0006,
"step": 5050
},
{
"epoch": 12.62,
"learning_rate": 1.816315789473684e-06,
"loss": 0.0007,
"step": 5075
},
{
"epoch": 12.69,
"learning_rate": 1.8071052631578945e-06,
"loss": 0.0007,
"step": 5100
},
{
"epoch": 12.75,
"learning_rate": 1.797894736842105e-06,
"loss": 0.0008,
"step": 5125
},
{
"epoch": 12.81,
"learning_rate": 1.7886842105263156e-06,
"loss": 0.0009,
"step": 5150
},
{
"epoch": 12.87,
"learning_rate": 1.7794736842105266e-06,
"loss": 0.0007,
"step": 5175
},
{
"epoch": 12.94,
"learning_rate": 1.7702631578947369e-06,
"loss": 0.0009,
"step": 5200
},
{
"epoch": 13.0,
"learning_rate": 1.7610526315789474e-06,
"loss": 0.0008,
"step": 5225
},
{
"epoch": 13.06,
"learning_rate": 1.751842105263158e-06,
"loss": 0.001,
"step": 5250
},
{
"epoch": 13.12,
"learning_rate": 1.7426315789473685e-06,
"loss": 0.0012,
"step": 5275
},
{
"epoch": 13.18,
"learning_rate": 1.733421052631579e-06,
"loss": 0.001,
"step": 5300
},
{
"epoch": 13.25,
"learning_rate": 1.7242105263157893e-06,
"loss": 0.001,
"step": 5325
},
{
"epoch": 13.31,
"learning_rate": 1.7149999999999999e-06,
"loss": 0.0007,
"step": 5350
},
{
"epoch": 13.37,
"learning_rate": 1.7057894736842104e-06,
"loss": 0.0008,
"step": 5375
},
{
"epoch": 13.43,
"learning_rate": 1.696578947368421e-06,
"loss": 0.0009,
"step": 5400
},
{
"epoch": 13.5,
"learning_rate": 1.6873684210526317e-06,
"loss": 0.0015,
"step": 5425
},
{
"epoch": 13.56,
"learning_rate": 1.678157894736842e-06,
"loss": 0.0011,
"step": 5450
},
{
"epoch": 13.62,
"learning_rate": 1.6689473684210526e-06,
"loss": 0.0013,
"step": 5475
},
{
"epoch": 13.68,
"learning_rate": 1.659736842105263e-06,
"loss": 0.0017,
"step": 5500
},
{
"epoch": 13.74,
"learning_rate": 1.6505263157894736e-06,
"loss": 0.0007,
"step": 5525
},
{
"epoch": 13.81,
"learning_rate": 1.6413157894736842e-06,
"loss": 0.0007,
"step": 5550
},
{
"epoch": 13.87,
"learning_rate": 1.6321052631578947e-06,
"loss": 0.0006,
"step": 5575
},
{
"epoch": 13.93,
"learning_rate": 1.6228947368421053e-06,
"loss": 0.0012,
"step": 5600
},
{
"epoch": 13.99,
"learning_rate": 1.6136842105263158e-06,
"loss": 0.0008,
"step": 5625
},
{
"epoch": 14.05,
"learning_rate": 1.6044736842105263e-06,
"loss": 0.0005,
"step": 5650
},
{
"epoch": 14.12,
"learning_rate": 1.5952631578947369e-06,
"loss": 0.0012,
"step": 5675
},
{
"epoch": 14.18,
"learning_rate": 1.5860526315789474e-06,
"loss": 0.0006,
"step": 5700
},
{
"epoch": 14.24,
"learning_rate": 1.5768421052631577e-06,
"loss": 0.0008,
"step": 5725
},
{
"epoch": 14.3,
"learning_rate": 1.5676315789473685e-06,
"loss": 0.0004,
"step": 5750
},
{
"epoch": 14.37,
"learning_rate": 1.558421052631579e-06,
"loss": 0.0004,
"step": 5775
},
{
"epoch": 14.43,
"learning_rate": 1.5492105263157895e-06,
"loss": 0.0004,
"step": 5800
},
{
"epoch": 14.49,
"learning_rate": 1.54e-06,
"loss": 0.0004,
"step": 5825
},
{
"epoch": 14.55,
"learning_rate": 1.5307894736842104e-06,
"loss": 0.0004,
"step": 5850
},
{
"epoch": 14.61,
"learning_rate": 1.521578947368421e-06,
"loss": 0.0006,
"step": 5875
},
{
"epoch": 14.68,
"learning_rate": 1.5123684210526315e-06,
"loss": 0.0005,
"step": 5900
},
{
"epoch": 14.74,
"learning_rate": 1.5031578947368422e-06,
"loss": 0.0004,
"step": 5925
},
{
"epoch": 14.8,
"learning_rate": 1.4939473684210528e-06,
"loss": 0.0003,
"step": 5950
},
{
"epoch": 14.86,
"learning_rate": 1.484736842105263e-06,
"loss": 0.0004,
"step": 5975
},
{
"epoch": 14.93,
"learning_rate": 1.4755263157894736e-06,
"loss": 0.0005,
"step": 6000
},
{
"epoch": 14.93,
"eval_loss": 0.34951019287109375,
"eval_runtime": 1981.6981,
"eval_samples_per_second": 0.856,
"eval_steps_per_second": 0.107,
"eval_wer": 10.196879643387815,
"step": 6000
},
{
"epoch": 14.99,
"learning_rate": 1.4663157894736842e-06,
"loss": 0.0005,
"step": 6025
},
{
"epoch": 15.05,
"learning_rate": 1.4571052631578947e-06,
"loss": 0.0004,
"step": 6050
},
{
"epoch": 15.11,
"learning_rate": 1.4478947368421052e-06,
"loss": 0.0005,
"step": 6075
},
{
"epoch": 15.17,
"learning_rate": 1.4386842105263158e-06,
"loss": 0.0005,
"step": 6100
},
{
"epoch": 15.24,
"learning_rate": 1.4294736842105263e-06,
"loss": 0.0004,
"step": 6125
},
{
"epoch": 15.3,
"learning_rate": 1.4202631578947368e-06,
"loss": 0.0006,
"step": 6150
},
{
"epoch": 15.36,
"learning_rate": 1.4110526315789474e-06,
"loss": 0.0005,
"step": 6175
},
{
"epoch": 15.42,
"learning_rate": 1.401842105263158e-06,
"loss": 0.0003,
"step": 6200
},
{
"epoch": 15.49,
"learning_rate": 1.3926315789473682e-06,
"loss": 0.0003,
"step": 6225
},
{
"epoch": 15.55,
"learning_rate": 1.3834210526315788e-06,
"loss": 0.0003,
"step": 6250
},
{
"epoch": 15.61,
"learning_rate": 1.3742105263157895e-06,
"loss": 0.0002,
"step": 6275
},
{
"epoch": 15.67,
"learning_rate": 1.365e-06,
"loss": 0.0002,
"step": 6300
},
{
"epoch": 15.73,
"learning_rate": 1.3557894736842106e-06,
"loss": 0.0003,
"step": 6325
},
{
"epoch": 15.8,
"learning_rate": 1.346578947368421e-06,
"loss": 0.0003,
"step": 6350
},
{
"epoch": 15.86,
"learning_rate": 1.3373684210526315e-06,
"loss": 0.0003,
"step": 6375
},
{
"epoch": 15.92,
"learning_rate": 1.328157894736842e-06,
"loss": 0.0003,
"step": 6400
},
{
"epoch": 15.98,
"learning_rate": 1.3189473684210525e-06,
"loss": 0.0004,
"step": 6425
},
{
"epoch": 16.04,
"learning_rate": 1.3097368421052633e-06,
"loss": 0.0002,
"step": 6450
},
{
"epoch": 16.11,
"learning_rate": 1.3005263157894736e-06,
"loss": 0.0004,
"step": 6475
},
{
"epoch": 16.17,
"learning_rate": 1.2913157894736841e-06,
"loss": 0.0003,
"step": 6500
},
{
"epoch": 16.23,
"learning_rate": 1.2821052631578947e-06,
"loss": 0.0002,
"step": 6525
},
{
"epoch": 16.29,
"learning_rate": 1.2728947368421052e-06,
"loss": 0.0002,
"step": 6550
},
{
"epoch": 16.36,
"learning_rate": 1.2636842105263157e-06,
"loss": 0.0002,
"step": 6575
},
{
"epoch": 16.42,
"learning_rate": 1.2544736842105263e-06,
"loss": 0.0002,
"step": 6600
},
{
"epoch": 16.48,
"learning_rate": 1.2452631578947368e-06,
"loss": 0.0002,
"step": 6625
},
{
"epoch": 16.54,
"learning_rate": 1.2360526315789474e-06,
"loss": 0.0002,
"step": 6650
},
{
"epoch": 16.6,
"learning_rate": 1.226842105263158e-06,
"loss": 0.0003,
"step": 6675
},
{
"epoch": 16.67,
"learning_rate": 1.2176315789473684e-06,
"loss": 0.0003,
"step": 6700
},
{
"epoch": 16.73,
"learning_rate": 1.208421052631579e-06,
"loss": 0.0002,
"step": 6725
},
{
"epoch": 16.79,
"learning_rate": 1.1992105263157893e-06,
"loss": 0.0003,
"step": 6750
},
{
"epoch": 16.85,
"learning_rate": 1.19e-06,
"loss": 0.0002,
"step": 6775
},
{
"epoch": 16.92,
"learning_rate": 1.1807894736842106e-06,
"loss": 0.0002,
"step": 6800
},
{
"epoch": 16.98,
"learning_rate": 1.1715789473684211e-06,
"loss": 0.0002,
"step": 6825
},
{
"epoch": 17.04,
"learning_rate": 1.1623684210526317e-06,
"loss": 0.0002,
"step": 6850
},
{
"epoch": 17.1,
"learning_rate": 1.153157894736842e-06,
"loss": 0.0002,
"step": 6875
},
{
"epoch": 17.16,
"learning_rate": 1.1439473684210525e-06,
"loss": 0.0002,
"step": 6900
},
{
"epoch": 17.23,
"learning_rate": 1.134736842105263e-06,
"loss": 0.0002,
"step": 6925
},
{
"epoch": 17.29,
"learning_rate": 1.1255263157894738e-06,
"loss": 0.0002,
"step": 6950
},
{
"epoch": 17.35,
"learning_rate": 1.1163157894736843e-06,
"loss": 0.0002,
"step": 6975
},
{
"epoch": 17.41,
"learning_rate": 1.1071052631578947e-06,
"loss": 0.0002,
"step": 7000
},
{
"epoch": 17.41,
"eval_loss": 0.3659239411354065,
"eval_runtime": 1978.6751,
"eval_samples_per_second": 0.857,
"eval_steps_per_second": 0.107,
"eval_wer": 10.25260029717682,
"step": 7000
},
{
"epoch": 17.48,
"learning_rate": 1.0978947368421052e-06,
"loss": 0.0002,
"step": 7025
},
{
"epoch": 17.54,
"learning_rate": 1.0886842105263157e-06,
"loss": 0.0002,
"step": 7050
},
{
"epoch": 17.6,
"learning_rate": 1.0794736842105263e-06,
"loss": 0.0002,
"step": 7075
},
{
"epoch": 17.66,
"learning_rate": 1.0702631578947368e-06,
"loss": 0.0002,
"step": 7100
},
{
"epoch": 17.72,
"learning_rate": 1.0610526315789473e-06,
"loss": 0.0002,
"step": 7125
},
{
"epoch": 17.79,
"learning_rate": 1.0518421052631579e-06,
"loss": 0.0002,
"step": 7150
},
{
"epoch": 17.85,
"learning_rate": 1.0426315789473684e-06,
"loss": 0.0002,
"step": 7175
},
{
"epoch": 17.91,
"learning_rate": 1.033421052631579e-06,
"loss": 0.0002,
"step": 7200
},
{
"epoch": 17.97,
"learning_rate": 1.0242105263157895e-06,
"loss": 0.0002,
"step": 7225
},
{
"epoch": 18.03,
"learning_rate": 1.0149999999999998e-06,
"loss": 0.0002,
"step": 7250
},
{
"epoch": 18.1,
"learning_rate": 1.0057894736842106e-06,
"loss": 0.0002,
"step": 7275
},
{
"epoch": 18.16,
"learning_rate": 9.96578947368421e-07,
"loss": 0.0002,
"step": 7300
},
{
"epoch": 18.22,
"learning_rate": 9.873684210526316e-07,
"loss": 0.0002,
"step": 7325
},
{
"epoch": 18.28,
"learning_rate": 9.781578947368422e-07,
"loss": 0.0002,
"step": 7350
},
{
"epoch": 18.35,
"learning_rate": 9.689473684210525e-07,
"loss": 0.0002,
"step": 7375
},
{
"epoch": 18.41,
"learning_rate": 9.59736842105263e-07,
"loss": 0.0002,
"step": 7400
},
{
"epoch": 18.47,
"learning_rate": 9.505263157894736e-07,
"loss": 0.0001,
"step": 7425
},
{
"epoch": 18.53,
"learning_rate": 9.413157894736843e-07,
"loss": 0.0002,
"step": 7450
},
{
"epoch": 18.59,
"learning_rate": 9.321052631578947e-07,
"loss": 0.0001,
"step": 7475
},
{
"epoch": 18.66,
"learning_rate": 9.228947368421053e-07,
"loss": 0.0001,
"step": 7500
},
{
"epoch": 18.72,
"learning_rate": 9.136842105263158e-07,
"loss": 0.0001,
"step": 7525
},
{
"epoch": 18.78,
"learning_rate": 9.044736842105262e-07,
"loss": 0.0001,
"step": 7550
},
{
"epoch": 18.84,
"learning_rate": 8.952631578947368e-07,
"loss": 0.0001,
"step": 7575
},
{
"epoch": 18.91,
"learning_rate": 8.860526315789473e-07,
"loss": 0.0001,
"step": 7600
},
{
"epoch": 18.97,
"learning_rate": 8.76842105263158e-07,
"loss": 0.0001,
"step": 7625
},
{
"epoch": 19.03,
"learning_rate": 8.676315789473684e-07,
"loss": 0.0001,
"step": 7650
},
{
"epoch": 19.09,
"learning_rate": 8.584210526315789e-07,
"loss": 0.0001,
"step": 7675
},
{
"epoch": 19.15,
"learning_rate": 8.492105263157895e-07,
"loss": 0.0001,
"step": 7700
},
{
"epoch": 19.22,
"learning_rate": 8.4e-07,
"loss": 0.0001,
"step": 7725
},
{
"epoch": 19.28,
"learning_rate": 8.307894736842105e-07,
"loss": 0.0001,
"step": 7750
},
{
"epoch": 19.34,
"learning_rate": 8.215789473684211e-07,
"loss": 0.0001,
"step": 7775
},
{
"epoch": 19.4,
"learning_rate": 8.123684210526315e-07,
"loss": 0.0001,
"step": 7800
},
{
"epoch": 19.47,
"learning_rate": 8.03157894736842e-07,
"loss": 0.0001,
"step": 7825
},
{
"epoch": 19.53,
"learning_rate": 7.939473684210527e-07,
"loss": 0.0001,
"step": 7850
},
{
"epoch": 19.59,
"learning_rate": 7.847368421052631e-07,
"loss": 0.0001,
"step": 7875
},
{
"epoch": 19.65,
"learning_rate": 7.755263157894737e-07,
"loss": 0.0001,
"step": 7900
},
{
"epoch": 19.71,
"learning_rate": 7.663157894736842e-07,
"loss": 0.0001,
"step": 7925
},
{
"epoch": 19.78,
"learning_rate": 7.571052631578947e-07,
"loss": 0.0001,
"step": 7950
},
{
"epoch": 19.84,
"learning_rate": 7.478947368421053e-07,
"loss": 0.0001,
"step": 7975
},
{
"epoch": 19.9,
"learning_rate": 7.386842105263158e-07,
"loss": 0.0001,
"step": 8000
},
{
"epoch": 19.9,
"eval_loss": 0.38460612297058105,
"eval_runtime": 1982.6401,
"eval_samples_per_second": 0.855,
"eval_steps_per_second": 0.107,
"eval_wer": 10.261887072808321,
"step": 8000
},
{
"epoch": 19.96,
"learning_rate": 7.294736842105263e-07,
"loss": 0.0001,
"step": 8025
},
{
"epoch": 20.02,
"learning_rate": 7.202631578947368e-07,
"loss": 0.0001,
"step": 8050
},
{
"epoch": 20.09,
"learning_rate": 7.110526315789473e-07,
"loss": 0.0001,
"step": 8075
},
{
"epoch": 20.15,
"learning_rate": 7.018421052631579e-07,
"loss": 0.0001,
"step": 8100
},
{
"epoch": 20.21,
"learning_rate": 6.926315789473684e-07,
"loss": 0.0001,
"step": 8125
},
{
"epoch": 20.27,
"learning_rate": 6.834210526315789e-07,
"loss": 0.0001,
"step": 8150
},
{
"epoch": 20.34,
"learning_rate": 6.742105263157894e-07,
"loss": 0.0001,
"step": 8175
},
{
"epoch": 20.4,
"learning_rate": 6.65e-07,
"loss": 0.0001,
"step": 8200
},
{
"epoch": 20.46,
"learning_rate": 6.557894736842105e-07,
"loss": 0.0001,
"step": 8225
},
{
"epoch": 20.52,
"learning_rate": 6.465789473684211e-07,
"loss": 0.0001,
"step": 8250
},
{
"epoch": 20.58,
"learning_rate": 6.373684210526316e-07,
"loss": 0.0001,
"step": 8275
},
{
"epoch": 20.65,
"learning_rate": 6.281578947368421e-07,
"loss": 0.0001,
"step": 8300
},
{
"epoch": 20.71,
"learning_rate": 6.189473684210526e-07,
"loss": 0.0001,
"step": 8325
},
{
"epoch": 20.77,
"learning_rate": 6.097368421052632e-07,
"loss": 0.0001,
"step": 8350
},
{
"epoch": 20.83,
"learning_rate": 6.005263157894736e-07,
"loss": 0.0001,
"step": 8375
},
{
"epoch": 20.9,
"learning_rate": 5.913157894736842e-07,
"loss": 0.0001,
"step": 8400
},
{
"epoch": 20.96,
"learning_rate": 5.821052631578948e-07,
"loss": 0.0001,
"step": 8425
},
{
"epoch": 21.02,
"learning_rate": 5.728947368421052e-07,
"loss": 0.0001,
"step": 8450
},
{
"epoch": 21.08,
"learning_rate": 5.636842105263158e-07,
"loss": 0.0001,
"step": 8475
},
{
"epoch": 21.14,
"learning_rate": 5.544736842105262e-07,
"loss": 0.0001,
"step": 8500
},
{
"epoch": 21.21,
"learning_rate": 5.452631578947369e-07,
"loss": 0.0001,
"step": 8525
},
{
"epoch": 21.27,
"learning_rate": 5.360526315789474e-07,
"loss": 0.0001,
"step": 8550
},
{
"epoch": 21.33,
"learning_rate": 5.268421052631578e-07,
"loss": 0.0001,
"step": 8575
},
{
"epoch": 21.39,
"learning_rate": 5.176315789473685e-07,
"loss": 0.0001,
"step": 8600
},
{
"epoch": 21.46,
"learning_rate": 5.084210526315789e-07,
"loss": 0.0001,
"step": 8625
},
{
"epoch": 21.52,
"learning_rate": 4.992105263157894e-07,
"loss": 0.0001,
"step": 8650
},
{
"epoch": 21.58,
"learning_rate": 4.900000000000001e-07,
"loss": 0.0001,
"step": 8675
},
{
"epoch": 21.64,
"learning_rate": 4.807894736842105e-07,
"loss": 0.0001,
"step": 8700
},
{
"epoch": 21.7,
"learning_rate": 4.7157894736842104e-07,
"loss": 0.0001,
"step": 8725
},
{
"epoch": 21.77,
"learning_rate": 4.623684210526315e-07,
"loss": 0.0001,
"step": 8750
},
{
"epoch": 21.83,
"learning_rate": 4.531578947368421e-07,
"loss": 0.0001,
"step": 8775
},
{
"epoch": 21.89,
"learning_rate": 4.4394736842105265e-07,
"loss": 0.0001,
"step": 8800
},
{
"epoch": 21.95,
"learning_rate": 4.347368421052632e-07,
"loss": 0.0001,
"step": 8825
},
{
"epoch": 22.01,
"learning_rate": 4.2552631578947367e-07,
"loss": 0.0001,
"step": 8850
},
{
"epoch": 22.08,
"learning_rate": 4.163157894736842e-07,
"loss": 0.0001,
"step": 8875
},
{
"epoch": 22.14,
"learning_rate": 4.071052631578947e-07,
"loss": 0.0001,
"step": 8900
},
{
"epoch": 22.2,
"learning_rate": 3.978947368421053e-07,
"loss": 0.0001,
"step": 8925
},
{
"epoch": 22.26,
"learning_rate": 3.886842105263158e-07,
"loss": 0.0001,
"step": 8950
},
{
"epoch": 22.33,
"learning_rate": 3.794736842105263e-07,
"loss": 0.0001,
"step": 8975
},
{
"epoch": 22.39,
"learning_rate": 3.7026315789473684e-07,
"loss": 0.0001,
"step": 9000
},
{
"epoch": 22.39,
"eval_loss": 0.3940609395503998,
"eval_runtime": 1985.3893,
"eval_samples_per_second": 0.854,
"eval_steps_per_second": 0.107,
"eval_wer": 10.289747399702822,
"step": 9000
},
{
"epoch": 22.45,
"learning_rate": 3.6105263157894737e-07,
"loss": 0.0001,
"step": 9025
},
{
"epoch": 22.51,
"learning_rate": 3.518421052631579e-07,
"loss": 0.0001,
"step": 9050
},
{
"epoch": 22.57,
"learning_rate": 3.426315789473684e-07,
"loss": 0.0001,
"step": 9075
},
{
"epoch": 22.64,
"learning_rate": 3.3342105263157893e-07,
"loss": 0.0001,
"step": 9100
},
{
"epoch": 22.7,
"learning_rate": 3.2421052631578947e-07,
"loss": 0.0001,
"step": 9125
},
{
"epoch": 22.76,
"learning_rate": 3.15e-07,
"loss": 0.0001,
"step": 9150
},
{
"epoch": 22.82,
"learning_rate": 3.0578947368421054e-07,
"loss": 0.0001,
"step": 9175
},
{
"epoch": 22.89,
"learning_rate": 2.96578947368421e-07,
"loss": 0.0001,
"step": 9200
},
{
"epoch": 22.95,
"learning_rate": 2.8736842105263156e-07,
"loss": 0.0001,
"step": 9225
},
{
"epoch": 23.01,
"learning_rate": 2.781578947368421e-07,
"loss": 0.0001,
"step": 9250
},
{
"epoch": 23.07,
"learning_rate": 2.6894736842105263e-07,
"loss": 0.0001,
"step": 9275
},
{
"epoch": 23.13,
"learning_rate": 2.5973684210526317e-07,
"loss": 0.0001,
"step": 9300
},
{
"epoch": 23.2,
"learning_rate": 2.5052631578947365e-07,
"loss": 0.0001,
"step": 9325
},
{
"epoch": 23.26,
"learning_rate": 2.413157894736842e-07,
"loss": 0.0001,
"step": 9350
},
{
"epoch": 23.32,
"learning_rate": 2.3210526315789475e-07,
"loss": 0.0001,
"step": 9375
},
{
"epoch": 23.38,
"learning_rate": 2.2289473684210526e-07,
"loss": 0.0001,
"step": 9400
},
{
"epoch": 23.45,
"learning_rate": 2.1368421052631577e-07,
"loss": 0.0001,
"step": 9425
},
{
"epoch": 23.51,
"learning_rate": 2.0447368421052633e-07,
"loss": 0.0001,
"step": 9450
},
{
"epoch": 23.57,
"learning_rate": 1.9526315789473684e-07,
"loss": 0.0001,
"step": 9475
},
{
"epoch": 23.63,
"learning_rate": 1.8605263157894735e-07,
"loss": 0.0001,
"step": 9500
},
{
"epoch": 23.69,
"learning_rate": 1.768421052631579e-07,
"loss": 0.0001,
"step": 9525
},
{
"epoch": 23.76,
"learning_rate": 1.676315789473684e-07,
"loss": 0.0001,
"step": 9550
},
{
"epoch": 23.82,
"learning_rate": 1.5842105263157896e-07,
"loss": 0.0001,
"step": 9575
},
{
"epoch": 23.88,
"learning_rate": 1.4921052631578947e-07,
"loss": 0.0001,
"step": 9600
},
{
"epoch": 23.94,
"learning_rate": 1.4e-07,
"loss": 0.0001,
"step": 9625
},
{
"epoch": 24.0,
"learning_rate": 1.3078947368421052e-07,
"loss": 0.0001,
"step": 9650
},
{
"epoch": 24.07,
"learning_rate": 1.2157894736842106e-07,
"loss": 0.0001,
"step": 9675
},
{
"epoch": 24.13,
"learning_rate": 1.1236842105263157e-07,
"loss": 0.0001,
"step": 9700
},
{
"epoch": 24.19,
"learning_rate": 1.031578947368421e-07,
"loss": 0.0001,
"step": 9725
},
{
"epoch": 24.25,
"learning_rate": 9.394736842105264e-08,
"loss": 0.0001,
"step": 9750
},
{
"epoch": 24.32,
"learning_rate": 8.473684210526316e-08,
"loss": 0.0001,
"step": 9775
},
{
"epoch": 24.38,
"learning_rate": 7.552631578947369e-08,
"loss": 0.0001,
"step": 9800
},
{
"epoch": 24.44,
"learning_rate": 6.631578947368421e-08,
"loss": 0.0001,
"step": 9825
},
{
"epoch": 24.5,
"learning_rate": 5.710526315789474e-08,
"loss": 0.0001,
"step": 9850
},
{
"epoch": 24.56,
"learning_rate": 4.789473684210526e-08,
"loss": 0.0001,
"step": 9875
},
{
"epoch": 24.63,
"learning_rate": 3.8684210526315785e-08,
"loss": 0.0001,
"step": 9900
},
{
"epoch": 24.69,
"learning_rate": 2.9473684210526315e-08,
"loss": 0.0001,
"step": 9925
},
{
"epoch": 24.75,
"learning_rate": 2.026315789473684e-08,
"loss": 0.0001,
"step": 9950
},
{
"epoch": 24.81,
"learning_rate": 1.1052631578947368e-08,
"loss": 0.0001,
"step": 9975
},
{
"epoch": 24.88,
"learning_rate": 1.8421052631578947e-09,
"loss": 0.0001,
"step": 10000
},
{
"epoch": 24.88,
"eval_loss": 0.3989764451980591,
"eval_runtime": 1984.699,
"eval_samples_per_second": 0.855,
"eval_steps_per_second": 0.107,
"eval_wer": 10.326894502228825,
"step": 10000
},
{
"epoch": 24.88,
"step": 10000,
"total_flos": 6.791089900290048e+20,
"train_loss": 0.013951553954230622,
"train_runtime": 87164.2955,
"train_samples_per_second": 3.671,
"train_steps_per_second": 0.115
}
],
"max_steps": 10000,
"num_train_epochs": 25,
"total_flos": 6.791089900290048e+20,
"trial_name": null,
"trial_params": null
}