xls-r-uyghur-cv8 / trainer_state.json
lucio's picture
End of training
233c538
raw
history blame
16.2 kB
{
"best_metric": null,
"best_model_checkpoint": null,
"epoch": 99.994708994709,
"global_step": 9400,
"is_hyper_param_search": false,
"is_local_process_zero": true,
"is_world_process_zero": true,
"log_history": [
{
"epoch": 1.06,
"learning_rate": 4.950000000000001e-06,
"loss": 13.5577,
"step": 100
},
{
"epoch": 2.13,
"learning_rate": 9.950000000000001e-06,
"loss": 5.868,
"step": 200
},
{
"epoch": 3.19,
"learning_rate": 1.4950000000000001e-05,
"loss": 4.1355,
"step": 300
},
{
"epoch": 4.25,
"learning_rate": 1.995e-05,
"loss": 3.5988,
"step": 400
},
{
"epoch": 5.32,
"learning_rate": 2.495e-05,
"loss": 3.3036,
"step": 500
},
{
"epoch": 5.32,
"eval_loss": 3.262770652770996,
"eval_runtime": 136.0729,
"eval_samples_per_second": 20.166,
"eval_steps_per_second": 2.521,
"eval_wer": 1.0,
"step": 500
},
{
"epoch": 6.38,
"learning_rate": 2.995e-05,
"loss": 3.2243,
"step": 600
},
{
"epoch": 7.44,
"learning_rate": 3.495e-05,
"loss": 3.1823,
"step": 700
},
{
"epoch": 8.51,
"learning_rate": 3.995e-05,
"loss": 3.1371,
"step": 800
},
{
"epoch": 9.57,
"learning_rate": 4.495e-05,
"loss": 3.0942,
"step": 900
},
{
"epoch": 10.63,
"learning_rate": 4.995e-05,
"loss": 2.9734,
"step": 1000
},
{
"epoch": 10.63,
"eval_loss": 2.5676724910736084,
"eval_runtime": 135.3086,
"eval_samples_per_second": 20.28,
"eval_steps_per_second": 2.535,
"eval_wer": 0.9980321266761246,
"step": 1000
},
{
"epoch": 11.7,
"learning_rate": 5.495e-05,
"loss": 2.2815,
"step": 1100
},
{
"epoch": 12.76,
"learning_rate": 5.995000000000001e-05,
"loss": 1.6698,
"step": 1200
},
{
"epoch": 13.83,
"learning_rate": 6.494999999999999e-05,
"loss": 1.4895,
"step": 1300
},
{
"epoch": 14.89,
"learning_rate": 6.995e-05,
"loss": 1.3959,
"step": 1400
},
{
"epoch": 15.95,
"learning_rate": 7.495e-05,
"loss": 1.3466,
"step": 1500
},
{
"epoch": 15.95,
"eval_loss": 0.44553351402282715,
"eval_runtime": 134.1872,
"eval_samples_per_second": 20.449,
"eval_steps_per_second": 2.556,
"eval_wer": 0.6306347535581895,
"step": 1500
},
{
"epoch": 17.02,
"learning_rate": 7.995e-05,
"loss": 1.3288,
"step": 1600
},
{
"epoch": 18.08,
"learning_rate": 8.495e-05,
"loss": 1.2866,
"step": 1700
},
{
"epoch": 19.15,
"learning_rate": 8.995e-05,
"loss": 1.2619,
"step": 1800
},
{
"epoch": 20.21,
"learning_rate": 9.495e-05,
"loss": 1.2496,
"step": 1900
},
{
"epoch": 21.28,
"learning_rate": 9.99e-05,
"loss": 1.2424,
"step": 2000
},
{
"epoch": 21.28,
"eval_loss": 0.3603059649467468,
"eval_runtime": 134.7251,
"eval_samples_per_second": 20.367,
"eval_steps_per_second": 2.546,
"eval_wer": 0.5301359205528351,
"step": 2000
},
{
"epoch": 22.34,
"learning_rate": 9.867567567567569e-05,
"loss": 1.2253,
"step": 2100
},
{
"epoch": 23.4,
"learning_rate": 9.732432432432433e-05,
"loss": 1.209,
"step": 2200
},
{
"epoch": 24.47,
"learning_rate": 9.597297297297298e-05,
"loss": 1.1984,
"step": 2300
},
{
"epoch": 25.53,
"learning_rate": 9.462162162162162e-05,
"loss": 1.1735,
"step": 2400
},
{
"epoch": 26.59,
"learning_rate": 9.327027027027028e-05,
"loss": 1.1655,
"step": 2500
},
{
"epoch": 26.59,
"eval_loss": 0.3164927661418915,
"eval_runtime": 135.1028,
"eval_samples_per_second": 20.31,
"eval_steps_per_second": 2.539,
"eval_wer": 0.4739828840785319,
"step": 2500
},
{
"epoch": 27.66,
"learning_rate": 9.191891891891893e-05,
"loss": 1.1529,
"step": 2600
},
{
"epoch": 28.72,
"learning_rate": 9.056756756756757e-05,
"loss": 1.1408,
"step": 2700
},
{
"epoch": 29.78,
"learning_rate": 8.921621621621622e-05,
"loss": 1.1201,
"step": 2800
},
{
"epoch": 30.85,
"learning_rate": 8.787837837837838e-05,
"loss": 1.1108,
"step": 2900
},
{
"epoch": 31.91,
"learning_rate": 8.652702702702703e-05,
"loss": 1.1026,
"step": 3000
},
{
"epoch": 31.91,
"eval_loss": 0.2930183410644531,
"eval_runtime": 135.7738,
"eval_samples_per_second": 20.21,
"eval_steps_per_second": 2.526,
"eval_wer": 0.4400256281177063,
"step": 3000
},
{
"epoch": 32.97,
"learning_rate": 8.517567567567568e-05,
"loss": 1.1035,
"step": 3100
},
{
"epoch": 34.04,
"learning_rate": 8.382432432432433e-05,
"loss": 1.0976,
"step": 3200
},
{
"epoch": 35.11,
"learning_rate": 8.247297297297298e-05,
"loss": 1.0717,
"step": 3300
},
{
"epoch": 36.17,
"learning_rate": 8.112162162162162e-05,
"loss": 1.0778,
"step": 3400
},
{
"epoch": 37.23,
"learning_rate": 7.977027027027028e-05,
"loss": 1.0655,
"step": 3500
},
{
"epoch": 37.23,
"eval_loss": 0.26754099130630493,
"eval_runtime": 134.1895,
"eval_samples_per_second": 20.449,
"eval_steps_per_second": 2.556,
"eval_wer": 0.41590773877625736,
"step": 3500
},
{
"epoch": 38.3,
"learning_rate": 7.841891891891892e-05,
"loss": 1.0588,
"step": 3600
},
{
"epoch": 39.36,
"learning_rate": 7.706756756756757e-05,
"loss": 1.0401,
"step": 3700
},
{
"epoch": 40.42,
"learning_rate": 7.571621621621621e-05,
"loss": 1.0296,
"step": 3800
},
{
"epoch": 41.49,
"learning_rate": 7.436486486486487e-05,
"loss": 1.0218,
"step": 3900
},
{
"epoch": 42.55,
"learning_rate": 7.301351351351352e-05,
"loss": 1.0239,
"step": 4000
},
{
"epoch": 42.55,
"eval_loss": 0.25800037384033203,
"eval_runtime": 134.1224,
"eval_samples_per_second": 20.459,
"eval_steps_per_second": 2.557,
"eval_wer": 0.3912864399798636,
"step": 4000
},
{
"epoch": 43.61,
"learning_rate": 7.166216216216216e-05,
"loss": 1.0226,
"step": 4100
},
{
"epoch": 44.68,
"learning_rate": 7.031081081081081e-05,
"loss": 1.0165,
"step": 4200
},
{
"epoch": 45.74,
"learning_rate": 6.895945945945947e-05,
"loss": 1.0083,
"step": 4300
},
{
"epoch": 46.8,
"learning_rate": 6.760810810810811e-05,
"loss": 0.9981,
"step": 4400
},
{
"epoch": 47.87,
"learning_rate": 6.625675675675676e-05,
"loss": 0.9938,
"step": 4500
},
{
"epoch": 47.87,
"eval_loss": 0.23732751607894897,
"eval_runtime": 135.9562,
"eval_samples_per_second": 20.183,
"eval_steps_per_second": 2.523,
"eval_wer": 0.36977712690494713,
"step": 4500
},
{
"epoch": 48.93,
"learning_rate": 6.49054054054054e-05,
"loss": 0.9717,
"step": 4600
},
{
"epoch": 49.99,
"learning_rate": 6.355405405405406e-05,
"loss": 0.9795,
"step": 4700
},
{
"epoch": 51.06,
"learning_rate": 6.22027027027027e-05,
"loss": 0.9804,
"step": 4800
},
{
"epoch": 52.13,
"learning_rate": 6.085135135135135e-05,
"loss": 0.9655,
"step": 4900
},
{
"epoch": 53.19,
"learning_rate": 5.95e-05,
"loss": 0.9655,
"step": 5000
},
{
"epoch": 53.19,
"eval_loss": 0.2379022240638733,
"eval_runtime": 134.3498,
"eval_samples_per_second": 20.424,
"eval_steps_per_second": 2.553,
"eval_wer": 0.3674889021097433,
"step": 5000
},
{
"epoch": 54.25,
"learning_rate": 5.8148648648648655e-05,
"loss": 0.9616,
"step": 5100
},
{
"epoch": 55.32,
"learning_rate": 5.67972972972973e-05,
"loss": 0.9633,
"step": 5200
},
{
"epoch": 56.38,
"learning_rate": 5.544594594594595e-05,
"loss": 0.94,
"step": 5300
},
{
"epoch": 57.44,
"learning_rate": 5.40945945945946e-05,
"loss": 0.9355,
"step": 5400
},
{
"epoch": 58.51,
"learning_rate": 5.274324324324325e-05,
"loss": 0.9374,
"step": 5500
},
{
"epoch": 58.51,
"eval_loss": 0.24859154224395752,
"eval_runtime": 135.3597,
"eval_samples_per_second": 20.272,
"eval_steps_per_second": 2.534,
"eval_wer": 0.3794792000366116,
"step": 5500
},
{
"epoch": 59.57,
"learning_rate": 5.1391891891891894e-05,
"loss": 0.93,
"step": 5600
},
{
"epoch": 60.63,
"learning_rate": 5.0040540540540546e-05,
"loss": 0.9212,
"step": 5700
},
{
"epoch": 61.7,
"learning_rate": 4.868918918918919e-05,
"loss": 0.9233,
"step": 5800
},
{
"epoch": 62.76,
"learning_rate": 4.733783783783784e-05,
"loss": 0.914,
"step": 5900
},
{
"epoch": 63.83,
"learning_rate": 4.598648648648649e-05,
"loss": 0.9065,
"step": 6000
},
{
"epoch": 63.83,
"eval_loss": 0.22428132593631744,
"eval_runtime": 136.639,
"eval_samples_per_second": 20.082,
"eval_steps_per_second": 2.51,
"eval_wer": 0.3405336140222415,
"step": 6000
},
{
"epoch": 64.89,
"learning_rate": 4.463513513513514e-05,
"loss": 0.9023,
"step": 6100
},
{
"epoch": 65.95,
"learning_rate": 4.3283783783783785e-05,
"loss": 0.9105,
"step": 6200
},
{
"epoch": 67.02,
"learning_rate": 4.193243243243244e-05,
"loss": 0.907,
"step": 6300
},
{
"epoch": 68.08,
"learning_rate": 4.058108108108108e-05,
"loss": 0.8939,
"step": 6400
},
{
"epoch": 69.15,
"learning_rate": 3.9229729729729734e-05,
"loss": 0.888,
"step": 6500
},
{
"epoch": 69.15,
"eval_loss": 0.21568605303764343,
"eval_runtime": 139.7942,
"eval_samples_per_second": 19.629,
"eval_steps_per_second": 2.454,
"eval_wer": 0.3277195551690998,
"step": 6500
},
{
"epoch": 70.21,
"learning_rate": 3.78918918918919e-05,
"loss": 0.8866,
"step": 6600
},
{
"epoch": 71.28,
"learning_rate": 3.654054054054054e-05,
"loss": 0.8881,
"step": 6700
},
{
"epoch": 72.34,
"learning_rate": 3.5189189189189195e-05,
"loss": 0.8849,
"step": 6800
},
{
"epoch": 73.4,
"learning_rate": 3.383783783783784e-05,
"loss": 0.8648,
"step": 6900
},
{
"epoch": 74.47,
"learning_rate": 3.2500000000000004e-05,
"loss": 0.8646,
"step": 7000
},
{
"epoch": 74.47,
"eval_loss": 0.21030458807945251,
"eval_runtime": 135.9762,
"eval_samples_per_second": 20.18,
"eval_steps_per_second": 2.522,
"eval_wer": 0.3287721385748936,
"step": 7000
},
{
"epoch": 75.53,
"learning_rate": 3.114864864864865e-05,
"loss": 0.8767,
"step": 7100
},
{
"epoch": 76.59,
"learning_rate": 2.97972972972973e-05,
"loss": 0.8655,
"step": 7200
},
{
"epoch": 77.66,
"learning_rate": 2.8445945945945946e-05,
"loss": 0.8672,
"step": 7300
},
{
"epoch": 78.72,
"learning_rate": 2.7094594594594598e-05,
"loss": 0.8625,
"step": 7400
},
{
"epoch": 79.78,
"learning_rate": 2.5743243243243243e-05,
"loss": 0.8602,
"step": 7500
},
{
"epoch": 79.78,
"eval_loss": 0.20880180597305298,
"eval_runtime": 135.4639,
"eval_samples_per_second": 20.256,
"eval_steps_per_second": 2.532,
"eval_wer": 0.32378380852134914,
"step": 7500
},
{
"epoch": 80.85,
"learning_rate": 2.4391891891891895e-05,
"loss": 0.8515,
"step": 7600
},
{
"epoch": 81.91,
"learning_rate": 2.3040540540540543e-05,
"loss": 0.8554,
"step": 7700
},
{
"epoch": 82.97,
"learning_rate": 2.1689189189189192e-05,
"loss": 0.8563,
"step": 7800
},
{
"epoch": 84.04,
"learning_rate": 2.033783783783784e-05,
"loss": 0.8655,
"step": 7900
},
{
"epoch": 85.11,
"learning_rate": 1.898648648648649e-05,
"loss": 0.8442,
"step": 8000
},
{
"epoch": 85.11,
"eval_loss": 0.2045026570558548,
"eval_runtime": 138.8412,
"eval_samples_per_second": 19.764,
"eval_steps_per_second": 2.47,
"eval_wer": 0.32657544277149786,
"step": 8000
},
{
"epoch": 86.17,
"learning_rate": 1.7635135135135137e-05,
"loss": 0.8323,
"step": 8100
},
{
"epoch": 87.23,
"learning_rate": 1.6283783783783786e-05,
"loss": 0.8384,
"step": 8200
},
{
"epoch": 88.3,
"learning_rate": 1.4932432432432433e-05,
"loss": 0.8391,
"step": 8300
},
{
"epoch": 89.36,
"learning_rate": 1.3581081081081081e-05,
"loss": 0.8292,
"step": 8400
},
{
"epoch": 90.42,
"learning_rate": 1.222972972972973e-05,
"loss": 0.8335,
"step": 8500
},
{
"epoch": 90.42,
"eval_loss": 0.20376762747764587,
"eval_runtime": 136.8377,
"eval_samples_per_second": 20.053,
"eval_steps_per_second": 2.507,
"eval_wer": 0.3240583954967736,
"step": 8500
},
{
"epoch": 91.49,
"learning_rate": 1.0878378378378378e-05,
"loss": 0.8314,
"step": 8600
},
{
"epoch": 92.55,
"learning_rate": 9.527027027027027e-06,
"loss": 0.8254,
"step": 8700
},
{
"epoch": 93.61,
"learning_rate": 8.175675675675675e-06,
"loss": 0.8231,
"step": 8800
},
{
"epoch": 94.68,
"learning_rate": 6.8243243243243244e-06,
"loss": 0.8164,
"step": 8900
},
{
"epoch": 95.74,
"learning_rate": 5.472972972972974e-06,
"loss": 0.8288,
"step": 9000
},
{
"epoch": 95.74,
"eval_loss": 0.20242640376091003,
"eval_runtime": 135.0305,
"eval_samples_per_second": 20.321,
"eval_steps_per_second": 2.54,
"eval_wer": 0.32799414214452427,
"step": 9000
},
{
"epoch": 96.8,
"learning_rate": 4.121621621621622e-06,
"loss": 0.816,
"step": 9100
},
{
"epoch": 97.87,
"learning_rate": 2.7702702702702708e-06,
"loss": 0.8163,
"step": 9200
},
{
"epoch": 98.93,
"learning_rate": 1.418918918918919e-06,
"loss": 0.8126,
"step": 9300
},
{
"epoch": 99.99,
"learning_rate": 6.756756756756757e-08,
"loss": 0.8084,
"step": 9400
},
{
"epoch": 99.99,
"step": 9400,
"total_flos": 1.0839754269306731e+20,
"train_loss": 1.3946523244330225,
"train_runtime": 41836.5581,
"train_samples_per_second": 14.423,
"train_steps_per_second": 0.225
}
],
"max_steps": 9400,
"num_train_epochs": 100,
"total_flos": 1.0839754269306731e+20,
"trial_name": null,
"trial_params": null
}