whisper_large_CGN / checkpoint-15000 /trainer_state.json
Jakob Poncelet
First model version
4a3d9de
raw
history blame
22.3 kB
{
"best_metric": 9.615871912312803,
"best_model_checkpoint": "/esat/audioslave/jponcele/whisper/finetuning_event/CGN/large/checkpoint-15000",
"epoch": 18.033533333333335,
"global_step": 15000,
"is_hyper_param_search": false,
"is_local_process_zero": true,
"is_world_process_zero": true,
"log_history": [
{
"epoch": 0.01,
"learning_rate": 1.94e-06,
"loss": 0.7504,
"step": 100
},
{
"epoch": 0.01,
"learning_rate": 3.920000000000001e-06,
"loss": 0.3565,
"step": 200
},
{
"epoch": 0.02,
"learning_rate": 5.92e-06,
"loss": 0.2271,
"step": 300
},
{
"epoch": 0.03,
"learning_rate": 7.92e-06,
"loss": 0.2206,
"step": 400
},
{
"epoch": 0.03,
"learning_rate": 9.920000000000002e-06,
"loss": 0.2448,
"step": 500
},
{
"epoch": 0.04,
"learning_rate": 9.933793103448277e-06,
"loss": 0.1832,
"step": 600
},
{
"epoch": 0.05,
"learning_rate": 9.865517241379312e-06,
"loss": 0.1504,
"step": 700
},
{
"epoch": 0.05,
"learning_rate": 9.796551724137931e-06,
"loss": 0.4013,
"step": 800
},
{
"epoch": 1.01,
"learning_rate": 9.727586206896552e-06,
"loss": 0.1325,
"step": 900
},
{
"epoch": 1.01,
"learning_rate": 9.658620689655173e-06,
"loss": 0.1078,
"step": 1000
},
{
"epoch": 1.01,
"eval_loss": 0.2717978060245514,
"eval_runtime": 90930.871,
"eval_samples_per_second": 0.568,
"eval_steps_per_second": 0.035,
"eval_wer": 14.3057057034114,
"step": 1000
},
{
"epoch": 1.02,
"learning_rate": 9.589655172413795e-06,
"loss": 0.1774,
"step": 1100
},
{
"epoch": 1.03,
"learning_rate": 9.520689655172414e-06,
"loss": 0.2996,
"step": 1200
},
{
"epoch": 1.03,
"learning_rate": 9.452413793103449e-06,
"loss": 0.1175,
"step": 1300
},
{
"epoch": 1.04,
"learning_rate": 9.38344827586207e-06,
"loss": 0.1589,
"step": 1400
},
{
"epoch": 1.05,
"learning_rate": 9.314482758620691e-06,
"loss": 0.2052,
"step": 1500
},
{
"epoch": 1.05,
"learning_rate": 9.24551724137931e-06,
"loss": 0.2372,
"step": 1600
},
{
"epoch": 2.01,
"learning_rate": 9.176551724137932e-06,
"loss": 0.1863,
"step": 1700
},
{
"epoch": 2.01,
"learning_rate": 9.107586206896553e-06,
"loss": 0.1293,
"step": 1800
},
{
"epoch": 2.02,
"learning_rate": 9.038620689655173e-06,
"loss": 0.1582,
"step": 1900
},
{
"epoch": 2.03,
"learning_rate": 8.969655172413794e-06,
"loss": 0.0762,
"step": 2000
},
{
"epoch": 2.03,
"eval_loss": 0.25997287034988403,
"eval_runtime": 86242.9732,
"eval_samples_per_second": 0.598,
"eval_steps_per_second": 0.037,
"eval_wer": 12.468659951078328,
"step": 2000
},
{
"epoch": 2.03,
"learning_rate": 8.900689655172415e-06,
"loss": 0.0869,
"step": 2100
},
{
"epoch": 2.04,
"learning_rate": 8.831724137931034e-06,
"loss": 0.067,
"step": 2200
},
{
"epoch": 2.05,
"learning_rate": 8.762758620689657e-06,
"loss": 0.2022,
"step": 2300
},
{
"epoch": 2.05,
"learning_rate": 8.693793103448277e-06,
"loss": 0.1808,
"step": 2400
},
{
"epoch": 3.01,
"learning_rate": 8.624827586206898e-06,
"loss": 0.1677,
"step": 2500
},
{
"epoch": 3.01,
"learning_rate": 8.555862068965517e-06,
"loss": 0.1633,
"step": 2600
},
{
"epoch": 3.02,
"learning_rate": 8.486896551724139e-06,
"loss": 0.1031,
"step": 2700
},
{
"epoch": 3.03,
"learning_rate": 8.41793103448276e-06,
"loss": 0.1452,
"step": 2800
},
{
"epoch": 3.03,
"learning_rate": 8.34896551724138e-06,
"loss": 0.0673,
"step": 2900
},
{
"epoch": 3.04,
"learning_rate": 8.28e-06,
"loss": 0.1941,
"step": 3000
},
{
"epoch": 3.04,
"eval_loss": 0.23860542476177216,
"eval_runtime": 87491.9677,
"eval_samples_per_second": 0.59,
"eval_steps_per_second": 0.037,
"eval_wer": 12.355587473912474,
"step": 3000
},
{
"epoch": 3.05,
"learning_rate": 8.211034482758622e-06,
"loss": 0.1519,
"step": 3100
},
{
"epoch": 3.05,
"learning_rate": 8.142068965517243e-06,
"loss": 0.1579,
"step": 3200
},
{
"epoch": 4.01,
"learning_rate": 8.073793103448276e-06,
"loss": 0.1228,
"step": 3300
},
{
"epoch": 4.01,
"learning_rate": 8.004827586206897e-06,
"loss": 0.2056,
"step": 3400
},
{
"epoch": 4.02,
"learning_rate": 7.935862068965518e-06,
"loss": 0.1543,
"step": 3500
},
{
"epoch": 4.03,
"learning_rate": 7.866896551724138e-06,
"loss": 0.0864,
"step": 3600
},
{
"epoch": 4.03,
"learning_rate": 7.797931034482759e-06,
"loss": 0.1363,
"step": 3700
},
{
"epoch": 4.04,
"learning_rate": 7.72896551724138e-06,
"loss": 0.0343,
"step": 3800
},
{
"epoch": 4.05,
"learning_rate": 7.660000000000001e-06,
"loss": 0.1021,
"step": 3900
},
{
"epoch": 4.05,
"learning_rate": 7.5910344827586215e-06,
"loss": 0.11,
"step": 4000
},
{
"epoch": 4.05,
"eval_loss": 0.24147863686084747,
"eval_runtime": 86034.4263,
"eval_samples_per_second": 0.6,
"eval_steps_per_second": 0.037,
"eval_wer": 11.524479299969311,
"step": 4000
},
{
"epoch": 5.0,
"learning_rate": 7.522068965517242e-06,
"loss": 0.0596,
"step": 4100
},
{
"epoch": 5.01,
"learning_rate": 7.453103448275862e-06,
"loss": 0.0889,
"step": 4200
},
{
"epoch": 5.02,
"learning_rate": 7.384137931034483e-06,
"loss": 0.0699,
"step": 4300
},
{
"epoch": 5.03,
"learning_rate": 7.315172413793104e-06,
"loss": 0.1914,
"step": 4400
},
{
"epoch": 5.03,
"learning_rate": 7.246206896551725e-06,
"loss": 0.1299,
"step": 4500
},
{
"epoch": 5.04,
"learning_rate": 7.177241379310346e-06,
"loss": 0.1244,
"step": 4600
},
{
"epoch": 5.04,
"learning_rate": 7.108275862068966e-06,
"loss": 0.079,
"step": 4700
},
{
"epoch": 5.05,
"learning_rate": 7.039310344827587e-06,
"loss": 0.0565,
"step": 4800
},
{
"epoch": 6.0,
"learning_rate": 6.970344827586207e-06,
"loss": 0.0424,
"step": 4900
},
{
"epoch": 6.01,
"learning_rate": 6.901379310344828e-06,
"loss": 0.0578,
"step": 5000
},
{
"epoch": 6.01,
"eval_loss": 0.23887480795383453,
"eval_runtime": 85622.6213,
"eval_samples_per_second": 0.603,
"eval_steps_per_second": 0.038,
"eval_wer": 11.166289121740093,
"step": 5000
},
{
"epoch": 6.02,
"learning_rate": 6.8324137931034485e-06,
"loss": 0.0527,
"step": 5100
},
{
"epoch": 6.02,
"learning_rate": 6.763448275862069e-06,
"loss": 0.0582,
"step": 5200
},
{
"epoch": 6.03,
"learning_rate": 6.694482758620691e-06,
"loss": 0.024,
"step": 5300
},
{
"epoch": 6.04,
"learning_rate": 6.625517241379311e-06,
"loss": 0.0562,
"step": 5400
},
{
"epoch": 6.04,
"learning_rate": 6.5565517241379315e-06,
"loss": 0.0607,
"step": 5500
},
{
"epoch": 6.05,
"learning_rate": 6.487586206896552e-06,
"loss": 0.0512,
"step": 5600
},
{
"epoch": 7.0,
"learning_rate": 6.418620689655173e-06,
"loss": 0.0434,
"step": 5700
},
{
"epoch": 7.01,
"learning_rate": 6.349655172413793e-06,
"loss": 0.0901,
"step": 5800
},
{
"epoch": 7.02,
"learning_rate": 6.280689655172414e-06,
"loss": 0.0687,
"step": 5900
},
{
"epoch": 7.02,
"learning_rate": 6.211724137931035e-06,
"loss": 0.0198,
"step": 6000
},
{
"epoch": 7.02,
"eval_loss": 0.2530022859573364,
"eval_runtime": 85352.5107,
"eval_samples_per_second": 0.605,
"eval_steps_per_second": 0.038,
"eval_wer": 10.562726574706145,
"step": 6000
},
{
"epoch": 7.03,
"learning_rate": 6.142758620689656e-06,
"loss": 0.0909,
"step": 6100
},
{
"epoch": 7.04,
"learning_rate": 6.073793103448276e-06,
"loss": 0.0906,
"step": 6200
},
{
"epoch": 7.04,
"learning_rate": 6.0048275862068975e-06,
"loss": 0.0599,
"step": 6300
},
{
"epoch": 7.05,
"learning_rate": 5.935862068965518e-06,
"loss": 0.0607,
"step": 6400
},
{
"epoch": 8.0,
"learning_rate": 5.866896551724138e-06,
"loss": 0.0932,
"step": 6500
},
{
"epoch": 8.01,
"learning_rate": 5.7979310344827585e-06,
"loss": 0.0902,
"step": 6600
},
{
"epoch": 8.02,
"learning_rate": 5.72896551724138e-06,
"loss": 0.0183,
"step": 6700
},
{
"epoch": 8.02,
"learning_rate": 5.66e-06,
"loss": 0.0264,
"step": 6800
},
{
"epoch": 8.03,
"learning_rate": 5.591034482758621e-06,
"loss": 0.0254,
"step": 6900
},
{
"epoch": 8.04,
"learning_rate": 5.522068965517242e-06,
"loss": 0.0924,
"step": 7000
},
{
"epoch": 8.04,
"eval_loss": 0.22727389633655548,
"eval_runtime": 85997.3481,
"eval_samples_per_second": 0.6,
"eval_steps_per_second": 0.038,
"eval_wer": 10.726350598532859,
"step": 7000
},
{
"epoch": 8.04,
"learning_rate": 5.453103448275863e-06,
"loss": 0.092,
"step": 7100
},
{
"epoch": 8.05,
"learning_rate": 5.384137931034483e-06,
"loss": 0.0653,
"step": 7200
},
{
"epoch": 9.0,
"learning_rate": 5.315172413793104e-06,
"loss": 0.0986,
"step": 7300
},
{
"epoch": 9.01,
"learning_rate": 5.2462068965517245e-06,
"loss": 0.1079,
"step": 7400
},
{
"epoch": 9.02,
"learning_rate": 5.177241379310345e-06,
"loss": 0.0404,
"step": 7500
},
{
"epoch": 9.02,
"learning_rate": 5.108275862068965e-06,
"loss": 0.0623,
"step": 7600
},
{
"epoch": 9.03,
"learning_rate": 5.039310344827587e-06,
"loss": 0.0219,
"step": 7700
},
{
"epoch": 9.04,
"learning_rate": 4.970344827586207e-06,
"loss": 0.0433,
"step": 7800
},
{
"epoch": 9.04,
"learning_rate": 4.901379310344828e-06,
"loss": 0.0555,
"step": 7900
},
{
"epoch": 9.05,
"learning_rate": 4.832413793103449e-06,
"loss": 0.0242,
"step": 8000
},
{
"epoch": 9.05,
"eval_loss": 0.2432514876127243,
"eval_runtime": 85600.137,
"eval_samples_per_second": 0.603,
"eval_steps_per_second": 0.038,
"eval_wer": 10.438448716920252,
"step": 8000
},
{
"epoch": 10.0,
"learning_rate": 4.763448275862069e-06,
"loss": 0.0146,
"step": 8100
},
{
"epoch": 10.01,
"learning_rate": 4.69448275862069e-06,
"loss": 0.0162,
"step": 8200
},
{
"epoch": 10.02,
"learning_rate": 4.625517241379311e-06,
"loss": 0.0422,
"step": 8300
},
{
"epoch": 10.02,
"learning_rate": 4.556551724137931e-06,
"loss": 0.0466,
"step": 8400
},
{
"epoch": 10.03,
"learning_rate": 4.487586206896552e-06,
"loss": 0.0264,
"step": 8500
},
{
"epoch": 10.04,
"learning_rate": 4.418620689655173e-06,
"loss": 0.0444,
"step": 8600
},
{
"epoch": 10.04,
"learning_rate": 4.349655172413794e-06,
"loss": 0.0785,
"step": 8700
},
{
"epoch": 10.05,
"learning_rate": 4.280689655172414e-06,
"loss": 0.0803,
"step": 8800
},
{
"epoch": 11.0,
"learning_rate": 4.2117241379310345e-06,
"loss": 0.0507,
"step": 8900
},
{
"epoch": 11.01,
"learning_rate": 4.142758620689656e-06,
"loss": 0.0468,
"step": 9000
},
{
"epoch": 11.01,
"eval_loss": 0.22253021597862244,
"eval_runtime": 85229.8997,
"eval_samples_per_second": 0.606,
"eval_steps_per_second": 0.038,
"eval_wer": 10.177159614280239,
"step": 9000
},
{
"epoch": 11.02,
"learning_rate": 4.073793103448276e-06,
"loss": 0.0714,
"step": 9100
},
{
"epoch": 11.02,
"learning_rate": 4.004827586206897e-06,
"loss": 0.0877,
"step": 9200
},
{
"epoch": 11.03,
"learning_rate": 3.9358620689655175e-06,
"loss": 0.0888,
"step": 9300
},
{
"epoch": 11.04,
"learning_rate": 3.867586206896552e-06,
"loss": 0.1097,
"step": 9400
},
{
"epoch": 11.04,
"learning_rate": 3.7986206896551727e-06,
"loss": 0.0995,
"step": 9500
},
{
"epoch": 11.05,
"learning_rate": 3.7296551724137935e-06,
"loss": 0.0074,
"step": 9600
},
{
"epoch": 12.0,
"learning_rate": 3.660689655172414e-06,
"loss": 0.0372,
"step": 9700
},
{
"epoch": 12.01,
"learning_rate": 3.5917241379310345e-06,
"loss": 0.028,
"step": 9800
},
{
"epoch": 12.02,
"learning_rate": 3.5227586206896553e-06,
"loss": 0.0528,
"step": 9900
},
{
"epoch": 12.02,
"learning_rate": 3.4537931034482765e-06,
"loss": 0.007,
"step": 10000
},
{
"epoch": 12.02,
"eval_loss": 0.2621050179004669,
"eval_runtime": 85409.6685,
"eval_samples_per_second": 0.604,
"eval_steps_per_second": 0.038,
"eval_wer": 9.971133411516329,
"step": 10000
},
{
"epoch": 12.03,
"learning_rate": 3.3848275862068968e-06,
"loss": 0.0174,
"step": 10100
},
{
"epoch": 12.04,
"learning_rate": 3.3158620689655175e-06,
"loss": 0.0164,
"step": 10200
},
{
"epoch": 12.04,
"learning_rate": 3.246896551724138e-06,
"loss": 0.0204,
"step": 10300
},
{
"epoch": 12.05,
"learning_rate": 3.177931034482759e-06,
"loss": 0.0057,
"step": 10400
},
{
"epoch": 13.0,
"learning_rate": 3.10896551724138e-06,
"loss": 0.0125,
"step": 10500
},
{
"epoch": 13.01,
"learning_rate": 3.04e-06,
"loss": 0.0346,
"step": 10600
},
{
"epoch": 13.02,
"learning_rate": 2.9717241379310345e-06,
"loss": 0.158,
"step": 10700
},
{
"epoch": 13.02,
"learning_rate": 2.9027586206896553e-06,
"loss": 0.1268,
"step": 10800
},
{
"epoch": 13.03,
"learning_rate": 2.8337931034482765e-06,
"loss": 0.0766,
"step": 10900
},
{
"epoch": 13.04,
"learning_rate": 2.764827586206897e-06,
"loss": 0.0103,
"step": 11000
},
{
"epoch": 13.04,
"eval_loss": 0.21942287683486938,
"eval_runtime": 85416.0012,
"eval_samples_per_second": 0.604,
"eval_steps_per_second": 0.038,
"eval_wer": 9.92351054388116,
"step": 11000
},
{
"epoch": 13.04,
"learning_rate": 2.6958620689655175e-06,
"loss": 0.0304,
"step": 11100
},
{
"epoch": 13.05,
"learning_rate": 2.626896551724138e-06,
"loss": 0.0132,
"step": 11200
},
{
"epoch": 14.0,
"learning_rate": 2.557931034482759e-06,
"loss": 0.0392,
"step": 11300
},
{
"epoch": 14.01,
"learning_rate": 2.4889655172413794e-06,
"loss": 0.0214,
"step": 11400
},
{
"epoch": 14.01,
"learning_rate": 2.42e-06,
"loss": 0.0431,
"step": 11500
},
{
"epoch": 14.02,
"learning_rate": 2.351034482758621e-06,
"loss": 0.0172,
"step": 11600
},
{
"epoch": 14.03,
"learning_rate": 2.2820689655172416e-06,
"loss": 0.0175,
"step": 11700
},
{
"epoch": 14.03,
"learning_rate": 2.2131034482758624e-06,
"loss": 0.0523,
"step": 11800
},
{
"epoch": 14.04,
"learning_rate": 2.1441379310344827e-06,
"loss": 0.089,
"step": 11900
},
{
"epoch": 14.05,
"learning_rate": 2.0758620689655175e-06,
"loss": 0.0571,
"step": 12000
},
{
"epoch": 14.05,
"eval_loss": 0.20878072082996368,
"eval_runtime": 128210.6226,
"eval_samples_per_second": 0.403,
"eval_steps_per_second": 0.025,
"eval_wer": 9.858697603703888,
"step": 12000
},
{
"epoch": 15.0,
"learning_rate": 2.0068965517241383e-06,
"loss": 0.0295,
"step": 12100
},
{
"epoch": 15.01,
"learning_rate": 1.9379310344827586e-06,
"loss": 0.0668,
"step": 12200
},
{
"epoch": 15.01,
"learning_rate": 1.8689655172413796e-06,
"loss": 0.0138,
"step": 12300
},
{
"epoch": 15.02,
"learning_rate": 1.8000000000000001e-06,
"loss": 0.0124,
"step": 12400
},
{
"epoch": 15.03,
"learning_rate": 1.7310344827586209e-06,
"loss": 0.0115,
"step": 12500
},
{
"epoch": 15.03,
"learning_rate": 1.6620689655172414e-06,
"loss": 0.0164,
"step": 12600
},
{
"epoch": 15.04,
"learning_rate": 1.5931034482758622e-06,
"loss": 0.0362,
"step": 12700
},
{
"epoch": 15.05,
"learning_rate": 1.524137931034483e-06,
"loss": 0.0362,
"step": 12800
},
{
"epoch": 16.0,
"learning_rate": 1.4551724137931037e-06,
"loss": 0.0596,
"step": 12900
},
{
"epoch": 16.01,
"learning_rate": 1.3862068965517242e-06,
"loss": 0.04,
"step": 13000
},
{
"epoch": 16.01,
"eval_loss": 0.21550078690052032,
"eval_runtime": 84922.7672,
"eval_samples_per_second": 0.608,
"eval_steps_per_second": 0.038,
"eval_wer": 9.765361876493786,
"step": 13000
},
{
"epoch": 16.01,
"learning_rate": 1.317241379310345e-06,
"loss": 0.0153,
"step": 13100
},
{
"epoch": 16.02,
"learning_rate": 1.2482758620689655e-06,
"loss": 0.015,
"step": 13200
},
{
"epoch": 16.03,
"learning_rate": 1.1793103448275863e-06,
"loss": 0.0159,
"step": 13300
},
{
"epoch": 16.03,
"learning_rate": 1.110344827586207e-06,
"loss": 0.0147,
"step": 13400
},
{
"epoch": 16.04,
"learning_rate": 1.0413793103448275e-06,
"loss": 0.0095,
"step": 13500
},
{
"epoch": 16.05,
"learning_rate": 9.724137931034483e-07,
"loss": 0.0058,
"step": 13600
},
{
"epoch": 17.0,
"learning_rate": 9.034482758620689e-07,
"loss": 0.0056,
"step": 13700
},
{
"epoch": 17.01,
"learning_rate": 8.344827586206897e-07,
"loss": 0.0159,
"step": 13800
},
{
"epoch": 17.01,
"learning_rate": 7.655172413793103e-07,
"loss": 0.0248,
"step": 13900
},
{
"epoch": 17.02,
"learning_rate": 6.965517241379311e-07,
"loss": 0.0191,
"step": 14000
},
{
"epoch": 17.02,
"eval_loss": 0.2494414746761322,
"eval_runtime": 84999.0835,
"eval_samples_per_second": 0.607,
"eval_steps_per_second": 0.038,
"eval_wer": 9.745243124925986,
"step": 14000
},
{
"epoch": 17.03,
"learning_rate": 6.275862068965517e-07,
"loss": 0.0314,
"step": 14100
},
{
"epoch": 17.03,
"learning_rate": 5.586206896551725e-07,
"loss": 0.0353,
"step": 14200
},
{
"epoch": 17.04,
"learning_rate": 4.896551724137931e-07,
"loss": 0.0035,
"step": 14300
},
{
"epoch": 17.05,
"learning_rate": 4.2068965517241383e-07,
"loss": 0.0112,
"step": 14400
},
{
"epoch": 18.0,
"learning_rate": 3.5172413793103453e-07,
"loss": 0.0039,
"step": 14500
},
{
"epoch": 18.01,
"learning_rate": 2.827586206896552e-07,
"loss": 0.0076,
"step": 14600
},
{
"epoch": 18.01,
"learning_rate": 2.1379310344827587e-07,
"loss": 0.0078,
"step": 14700
},
{
"epoch": 18.02,
"learning_rate": 1.4482758620689657e-07,
"loss": 0.0141,
"step": 14800
},
{
"epoch": 18.03,
"learning_rate": 7.586206896551724e-08,
"loss": 0.0125,
"step": 14900
},
{
"epoch": 18.03,
"learning_rate": 6.896551724137931e-09,
"loss": 0.0149,
"step": 15000
},
{
"epoch": 18.03,
"eval_loss": 0.23932012915611267,
"eval_runtime": 85191.5609,
"eval_samples_per_second": 0.606,
"eval_steps_per_second": 0.038,
"eval_wer": 9.615871912312803,
"step": 15000
}
],
"max_steps": 15000,
"num_train_epochs": 9223372036854775807,
"total_flos": 2.0390255069036544e+21,
"trial_name": null,
"trial_params": null
}