xls-r-uyghur-cv8 / trainer_state.json
lucio's picture
End of training
c81d5d1
raw
history blame
16.3 kB
{
"best_metric": null,
"best_model_checkpoint": null,
"epoch": 49.99602649006623,
"global_step": 9400,
"is_hyper_param_search": false,
"is_local_process_zero": true,
"is_world_process_zero": true,
"log_history": [
{
"epoch": 0.53,
"learning_rate": 3.7125e-06,
"loss": 15.287,
"step": 100
},
{
"epoch": 1.06,
"learning_rate": 7.4625e-06,
"loss": 7.6558,
"step": 200
},
{
"epoch": 1.59,
"learning_rate": 1.1212499999999998e-05,
"loss": 4.9409,
"step": 300
},
{
"epoch": 2.13,
"learning_rate": 1.49625e-05,
"loss": 4.2283,
"step": 400
},
{
"epoch": 2.66,
"learning_rate": 1.8712499999999997e-05,
"loss": 3.6871,
"step": 500
},
{
"epoch": 2.66,
"eval_loss": 3.5374293327331543,
"eval_runtime": 140.96,
"eval_samples_per_second": 19.467,
"eval_steps_per_second": 2.433,
"eval_wer": 1.0,
"step": 500
},
{
"epoch": 3.19,
"learning_rate": 2.2462499999999997e-05,
"loss": 3.4073,
"step": 600
},
{
"epoch": 3.72,
"learning_rate": 2.6212499999999997e-05,
"loss": 3.2613,
"step": 700
},
{
"epoch": 4.25,
"learning_rate": 2.99625e-05,
"loss": 3.2504,
"step": 800
},
{
"epoch": 4.78,
"learning_rate": 3.37125e-05,
"loss": 3.1732,
"step": 900
},
{
"epoch": 5.32,
"learning_rate": 3.7462499999999996e-05,
"loss": 3.1501,
"step": 1000
},
{
"epoch": 5.32,
"eval_loss": 3.127795934677124,
"eval_runtime": 140.7753,
"eval_samples_per_second": 19.492,
"eval_steps_per_second": 2.437,
"eval_wer": 1.0,
"step": 1000
},
{
"epoch": 5.85,
"learning_rate": 4.12125e-05,
"loss": 3.0666,
"step": 1100
},
{
"epoch": 6.38,
"learning_rate": 4.4962499999999995e-05,
"loss": 2.7047,
"step": 1200
},
{
"epoch": 6.91,
"learning_rate": 4.871249999999999e-05,
"loss": 2.0895,
"step": 1300
},
{
"epoch": 7.45,
"learning_rate": 5.2462499999999994e-05,
"loss": 1.7459,
"step": 1400
},
{
"epoch": 7.97,
"learning_rate": 5.62125e-05,
"loss": 1.5843,
"step": 1500
},
{
"epoch": 7.97,
"eval_loss": 0.6358404755592346,
"eval_runtime": 138.8212,
"eval_samples_per_second": 19.766,
"eval_steps_per_second": 2.471,
"eval_wer": 0.6914246512691516,
"step": 1500
},
{
"epoch": 8.51,
"learning_rate": 5.9962499999999994e-05,
"loss": 1.5166,
"step": 1600
},
{
"epoch": 9.04,
"learning_rate": 6.37125e-05,
"loss": 1.4555,
"step": 1700
},
{
"epoch": 9.57,
"learning_rate": 6.746249999999999e-05,
"loss": 1.4013,
"step": 1800
},
{
"epoch": 10.11,
"learning_rate": 7.121249999999999e-05,
"loss": 1.3841,
"step": 1900
},
{
"epoch": 10.64,
"learning_rate": 7.49625e-05,
"loss": 1.3378,
"step": 2000
},
{
"epoch": 10.64,
"eval_loss": 0.442169189453125,
"eval_runtime": 140.0078,
"eval_samples_per_second": 19.599,
"eval_steps_per_second": 2.45,
"eval_wer": 0.5924536931168534,
"step": 2000
},
{
"epoch": 11.17,
"learning_rate": 7.400675675675676e-05,
"loss": 1.3249,
"step": 2100
},
{
"epoch": 11.7,
"learning_rate": 7.299324324324324e-05,
"loss": 1.2996,
"step": 2200
},
{
"epoch": 12.23,
"learning_rate": 7.197972972972972e-05,
"loss": 1.2952,
"step": 2300
},
{
"epoch": 12.76,
"learning_rate": 7.096621621621621e-05,
"loss": 1.2626,
"step": 2400
},
{
"epoch": 13.3,
"learning_rate": 6.99527027027027e-05,
"loss": 1.2595,
"step": 2500
},
{
"epoch": 13.3,
"eval_loss": 0.3921487033367157,
"eval_runtime": 140.1536,
"eval_samples_per_second": 19.579,
"eval_steps_per_second": 2.447,
"eval_wer": 0.5511548136290876,
"step": 2500
},
{
"epoch": 13.83,
"learning_rate": 6.893918918918919e-05,
"loss": 1.2282,
"step": 2600
},
{
"epoch": 14.36,
"learning_rate": 6.792567567567567e-05,
"loss": 1.2352,
"step": 2700
},
{
"epoch": 14.89,
"learning_rate": 6.691216216216216e-05,
"loss": 1.2129,
"step": 2800
},
{
"epoch": 15.42,
"learning_rate": 6.589864864864864e-05,
"loss": 1.2184,
"step": 2900
},
{
"epoch": 15.95,
"learning_rate": 6.488513513513514e-05,
"loss": 1.1643,
"step": 3000
},
{
"epoch": 15.95,
"eval_loss": 0.35074228048324585,
"eval_runtime": 142.5452,
"eval_samples_per_second": 19.25,
"eval_steps_per_second": 2.406,
"eval_wer": 0.514932540589984,
"step": 3000
},
{
"epoch": 16.49,
"learning_rate": 6.388175675675675e-05,
"loss": 1.1704,
"step": 3100
},
{
"epoch": 17.02,
"learning_rate": 6.287837837837837e-05,
"loss": 1.1774,
"step": 3200
},
{
"epoch": 17.55,
"learning_rate": 6.186486486486485e-05,
"loss": 1.157,
"step": 3300
},
{
"epoch": 18.08,
"learning_rate": 6.085135135135135e-05,
"loss": 1.1605,
"step": 3400
},
{
"epoch": 18.61,
"learning_rate": 5.983783783783783e-05,
"loss": 1.1352,
"step": 3500
},
{
"epoch": 18.61,
"eval_loss": 0.3350585401058197,
"eval_runtime": 140.4732,
"eval_samples_per_second": 19.534,
"eval_steps_per_second": 2.442,
"eval_wer": 0.5018980105190944,
"step": 3500
},
{
"epoch": 19.15,
"learning_rate": 5.882432432432432e-05,
"loss": 1.1474,
"step": 3600
},
{
"epoch": 19.68,
"learning_rate": 5.781081081081081e-05,
"loss": 1.1273,
"step": 3700
},
{
"epoch": 20.21,
"learning_rate": 5.679729729729729e-05,
"loss": 1.133,
"step": 3800
},
{
"epoch": 20.74,
"learning_rate": 5.578378378378378e-05,
"loss": 1.1112,
"step": 3900
},
{
"epoch": 21.28,
"learning_rate": 5.477027027027026e-05,
"loss": 1.1113,
"step": 4000
},
{
"epoch": 21.28,
"eval_loss": 0.3152759373188019,
"eval_runtime": 139.3947,
"eval_samples_per_second": 19.685,
"eval_steps_per_second": 2.461,
"eval_wer": 0.4845186370912417,
"step": 4000
},
{
"epoch": 21.81,
"learning_rate": 5.375675675675675e-05,
"loss": 1.1029,
"step": 4100
},
{
"epoch": 22.34,
"learning_rate": 5.274324324324324e-05,
"loss": 1.1124,
"step": 4200
},
{
"epoch": 22.87,
"learning_rate": 5.172972972972972e-05,
"loss": 1.091,
"step": 4300
},
{
"epoch": 23.4,
"learning_rate": 5.071621621621621e-05,
"loss": 1.097,
"step": 4400
},
{
"epoch": 23.93,
"learning_rate": 4.97027027027027e-05,
"loss": 1.0914,
"step": 4500
},
{
"epoch": 23.93,
"eval_loss": 0.3050296902656555,
"eval_runtime": 137.2603,
"eval_samples_per_second": 19.991,
"eval_steps_per_second": 2.499,
"eval_wer": 0.4594100160073176,
"step": 4500
},
{
"epoch": 24.47,
"learning_rate": 4.8689189189189184e-05,
"loss": 1.0749,
"step": 4600
},
{
"epoch": 25.0,
"learning_rate": 4.767567567567567e-05,
"loss": 1.0807,
"step": 4700
},
{
"epoch": 25.53,
"learning_rate": 4.666216216216216e-05,
"loss": 1.0647,
"step": 4800
},
{
"epoch": 26.06,
"learning_rate": 4.564864864864864e-05,
"loss": 1.068,
"step": 4900
},
{
"epoch": 26.59,
"learning_rate": 4.463513513513513e-05,
"loss": 1.0468,
"step": 5000
},
{
"epoch": 26.59,
"eval_loss": 0.2889558672904968,
"eval_runtime": 137.6066,
"eval_samples_per_second": 19.941,
"eval_steps_per_second": 2.493,
"eval_wer": 0.44701577864166475,
"step": 5000
},
{
"epoch": 27.13,
"learning_rate": 4.3621621621621624e-05,
"loss": 1.0548,
"step": 5100
},
{
"epoch": 27.66,
"learning_rate": 4.26081081081081e-05,
"loss": 1.0423,
"step": 5200
},
{
"epoch": 28.19,
"learning_rate": 4.160472972972973e-05,
"loss": 1.0436,
"step": 5300
},
{
"epoch": 28.72,
"learning_rate": 4.059121621621621e-05,
"loss": 1.0338,
"step": 5400
},
{
"epoch": 29.25,
"learning_rate": 3.95777027027027e-05,
"loss": 1.0473,
"step": 5500
},
{
"epoch": 29.25,
"eval_loss": 0.27552109956741333,
"eval_runtime": 139.7371,
"eval_samples_per_second": 19.637,
"eval_steps_per_second": 2.455,
"eval_wer": 0.43306654470615136,
"step": 5500
},
{
"epoch": 29.78,
"learning_rate": 3.856418918918919e-05,
"loss": 1.027,
"step": 5600
},
{
"epoch": 30.32,
"learning_rate": 3.755067567567568e-05,
"loss": 1.0418,
"step": 5700
},
{
"epoch": 30.85,
"learning_rate": 3.653716216216216e-05,
"loss": 1.0147,
"step": 5800
},
{
"epoch": 31.38,
"learning_rate": 3.552364864864865e-05,
"loss": 1.0197,
"step": 5900
},
{
"epoch": 31.91,
"learning_rate": 3.451013513513513e-05,
"loss": 1.0065,
"step": 6000
},
{
"epoch": 31.91,
"eval_loss": 0.27181389927864075,
"eval_runtime": 138.4335,
"eval_samples_per_second": 19.822,
"eval_steps_per_second": 2.478,
"eval_wer": 0.42638920649439743,
"step": 6000
},
{
"epoch": 32.45,
"learning_rate": 3.3496621621621615e-05,
"loss": 1.0138,
"step": 6100
},
{
"epoch": 32.97,
"learning_rate": 3.2483108108108105e-05,
"loss": 1.0027,
"step": 6200
},
{
"epoch": 33.51,
"learning_rate": 3.146959459459459e-05,
"loss": 1.0157,
"step": 6300
},
{
"epoch": 34.04,
"learning_rate": 3.045608108108108e-05,
"loss": 1.0034,
"step": 6400
},
{
"epoch": 34.57,
"learning_rate": 2.9442567567567563e-05,
"loss": 0.9794,
"step": 6500
},
{
"epoch": 34.57,
"eval_loss": 0.2646064758300781,
"eval_runtime": 138.3623,
"eval_samples_per_second": 19.832,
"eval_steps_per_second": 2.479,
"eval_wer": 0.41934598673679396,
"step": 6500
},
{
"epoch": 35.11,
"learning_rate": 2.8429054054054054e-05,
"loss": 0.9911,
"step": 6600
},
{
"epoch": 35.64,
"learning_rate": 2.7415540540540538e-05,
"loss": 0.9807,
"step": 6700
},
{
"epoch": 36.17,
"learning_rate": 2.6402027027027025e-05,
"loss": 0.9843,
"step": 6800
},
{
"epoch": 36.7,
"learning_rate": 2.538851351351351e-05,
"loss": 0.9735,
"step": 6900
},
{
"epoch": 37.23,
"learning_rate": 2.4375e-05,
"loss": 0.9849,
"step": 7000
},
{
"epoch": 37.23,
"eval_loss": 0.2609545886516571,
"eval_runtime": 144.7382,
"eval_samples_per_second": 18.958,
"eval_steps_per_second": 2.37,
"eval_wer": 0.4058083695403613,
"step": 7000
},
{
"epoch": 37.76,
"learning_rate": 2.3361486486486483e-05,
"loss": 0.983,
"step": 7100
},
{
"epoch": 38.3,
"learning_rate": 2.234797297297297e-05,
"loss": 0.9775,
"step": 7200
},
{
"epoch": 38.83,
"learning_rate": 2.1334459459459458e-05,
"loss": 0.9597,
"step": 7300
},
{
"epoch": 39.36,
"learning_rate": 2.0320945945945945e-05,
"loss": 0.9606,
"step": 7400
},
{
"epoch": 39.89,
"learning_rate": 1.930743243243243e-05,
"loss": 0.9496,
"step": 7500
},
{
"epoch": 39.89,
"eval_loss": 0.2522386610507965,
"eval_runtime": 139.7036,
"eval_samples_per_second": 19.642,
"eval_steps_per_second": 2.455,
"eval_wer": 0.3984907386233707,
"step": 7500
},
{
"epoch": 40.42,
"learning_rate": 1.829391891891892e-05,
"loss": 0.9581,
"step": 7600
},
{
"epoch": 40.95,
"learning_rate": 1.7280405405405403e-05,
"loss": 0.9477,
"step": 7700
},
{
"epoch": 41.49,
"learning_rate": 1.626689189189189e-05,
"loss": 0.9538,
"step": 7800
},
{
"epoch": 42.02,
"learning_rate": 1.5253378378378378e-05,
"loss": 0.9631,
"step": 7900
},
{
"epoch": 42.55,
"learning_rate": 1.4239864864864863e-05,
"loss": 0.9367,
"step": 8000
},
{
"epoch": 42.55,
"eval_loss": 0.25142449140548706,
"eval_runtime": 138.6762,
"eval_samples_per_second": 19.787,
"eval_steps_per_second": 2.473,
"eval_wer": 0.3946947175851818,
"step": 8000
},
{
"epoch": 43.08,
"learning_rate": 1.322635135135135e-05,
"loss": 0.9498,
"step": 8100
},
{
"epoch": 43.61,
"learning_rate": 1.2212837837837838e-05,
"loss": 0.9389,
"step": 8200
},
{
"epoch": 44.15,
"learning_rate": 1.1199324324324323e-05,
"loss": 0.937,
"step": 8300
},
{
"epoch": 44.68,
"learning_rate": 1.0195945945945945e-05,
"loss": 0.9394,
"step": 8400
},
{
"epoch": 45.21,
"learning_rate": 9.182432432432432e-06,
"loss": 0.9295,
"step": 8500
},
{
"epoch": 45.21,
"eval_loss": 0.24582630395889282,
"eval_runtime": 140.7119,
"eval_samples_per_second": 19.501,
"eval_steps_per_second": 2.438,
"eval_wer": 0.3883375257260462,
"step": 8500
},
{
"epoch": 45.74,
"learning_rate": 8.168918918918917e-06,
"loss": 0.9244,
"step": 8600
},
{
"epoch": 46.28,
"learning_rate": 7.1554054054054045e-06,
"loss": 0.9273,
"step": 8700
},
{
"epoch": 46.81,
"learning_rate": 6.141891891891891e-06,
"loss": 0.9229,
"step": 8800
},
{
"epoch": 47.34,
"learning_rate": 5.128378378378377e-06,
"loss": 0.9392,
"step": 8900
},
{
"epoch": 47.87,
"learning_rate": 4.1148648648648645e-06,
"loss": 0.9187,
"step": 9000
},
{
"epoch": 47.87,
"eval_loss": 0.24390903115272522,
"eval_runtime": 139.6323,
"eval_samples_per_second": 19.652,
"eval_steps_per_second": 2.456,
"eval_wer": 0.3832609192773839,
"step": 9000
},
{
"epoch": 48.4,
"learning_rate": 3.1013513513513513e-06,
"loss": 0.9194,
"step": 9100
},
{
"epoch": 48.93,
"learning_rate": 2.0878378378378376e-06,
"loss": 0.9234,
"step": 9200
},
{
"epoch": 49.47,
"learning_rate": 1.0743243243243242e-06,
"loss": 0.9257,
"step": 9300
},
{
"epoch": 50.0,
"learning_rate": 6.08108108108108e-08,
"loss": 0.9097,
"step": 9400
},
{
"epoch": 50.0,
"step": 9400,
"total_flos": 5.3948756860365595e+19,
"train_loss": 1.5738848333155855,
"train_runtime": 24806.553,
"train_samples_per_second": 12.162,
"train_steps_per_second": 0.379
}
],
"max_steps": 9400,
"num_train_epochs": 50,
"total_flos": 5.3948756860365595e+19,
"trial_name": null,
"trial_params": null
}