whisper_small_CGN / checkpoint-15000 /trainer_state.json
Jakob Poncelet
First model version
33fd6fc
raw
history blame
22.3 kB
{
"best_metric": 15.197170132057957,
"best_model_checkpoint": "/esat/audioslave/jponcele/whisper/finetuning_event/CGN/small/checkpoint-14000",
"epoch": 18.0334,
"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": 1.98e-06,
"loss": 1.2858,
"step": 100
},
{
"epoch": 0.01,
"learning_rate": 3.980000000000001e-06,
"loss": 0.6674,
"step": 200
},
{
"epoch": 0.02,
"learning_rate": 5.98e-06,
"loss": 0.4315,
"step": 300
},
{
"epoch": 0.03,
"learning_rate": 7.980000000000002e-06,
"loss": 0.4035,
"step": 400
},
{
"epoch": 0.03,
"learning_rate": 9.980000000000001e-06,
"loss": 0.4319,
"step": 500
},
{
"epoch": 0.04,
"learning_rate": 9.931724137931036e-06,
"loss": 0.304,
"step": 600
},
{
"epoch": 0.05,
"learning_rate": 9.862758620689657e-06,
"loss": 0.2622,
"step": 700
},
{
"epoch": 0.05,
"learning_rate": 9.793793103448276e-06,
"loss": 0.613,
"step": 800
},
{
"epoch": 1.01,
"learning_rate": 9.724827586206897e-06,
"loss": 0.2377,
"step": 900
},
{
"epoch": 1.01,
"learning_rate": 9.655862068965519e-06,
"loss": 0.1967,
"step": 1000
},
{
"epoch": 1.01,
"eval_loss": 0.4084566831588745,
"eval_runtime": 38894.1602,
"eval_samples_per_second": 1.327,
"eval_steps_per_second": 0.166,
"eval_wer": 21.84590818973256,
"step": 1000
},
{
"epoch": 1.02,
"learning_rate": 9.586896551724138e-06,
"loss": 0.2913,
"step": 1100
},
{
"epoch": 1.03,
"learning_rate": 9.51793103448276e-06,
"loss": 0.4736,
"step": 1200
},
{
"epoch": 1.03,
"learning_rate": 9.44896551724138e-06,
"loss": 0.2149,
"step": 1300
},
{
"epoch": 1.04,
"learning_rate": 9.38e-06,
"loss": 0.2689,
"step": 1400
},
{
"epoch": 1.05,
"learning_rate": 9.311034482758621e-06,
"loss": 0.3469,
"step": 1500
},
{
"epoch": 1.05,
"learning_rate": 9.242068965517242e-06,
"loss": 0.3798,
"step": 1600
},
{
"epoch": 2.01,
"learning_rate": 9.173103448275864e-06,
"loss": 0.3055,
"step": 1700
},
{
"epoch": 2.01,
"learning_rate": 9.104137931034483e-06,
"loss": 0.2221,
"step": 1800
},
{
"epoch": 2.02,
"learning_rate": 9.035172413793104e-06,
"loss": 0.2692,
"step": 1900
},
{
"epoch": 2.03,
"learning_rate": 8.966206896551725e-06,
"loss": 0.1355,
"step": 2000
},
{
"epoch": 2.03,
"eval_loss": 0.3751971423625946,
"eval_runtime": 37321.448,
"eval_samples_per_second": 1.383,
"eval_steps_per_second": 0.173,
"eval_wer": 18.621177914704134,
"step": 2000
},
{
"epoch": 2.03,
"learning_rate": 8.897241379310345e-06,
"loss": 0.1604,
"step": 2100
},
{
"epoch": 2.04,
"learning_rate": 8.828275862068966e-06,
"loss": 0.1269,
"step": 2200
},
{
"epoch": 2.05,
"learning_rate": 8.759310344827587e-06,
"loss": 0.3063,
"step": 2300
},
{
"epoch": 2.05,
"learning_rate": 8.690344827586208e-06,
"loss": 0.2932,
"step": 2400
},
{
"epoch": 3.01,
"learning_rate": 8.621379310344828e-06,
"loss": 0.2635,
"step": 2500
},
{
"epoch": 3.01,
"learning_rate": 8.552413793103449e-06,
"loss": 0.2541,
"step": 2600
},
{
"epoch": 3.02,
"learning_rate": 8.48344827586207e-06,
"loss": 0.1753,
"step": 2700
},
{
"epoch": 3.03,
"learning_rate": 8.41448275862069e-06,
"loss": 0.2342,
"step": 2800
},
{
"epoch": 3.03,
"learning_rate": 8.34551724137931e-06,
"loss": 0.1209,
"step": 2900
},
{
"epoch": 3.04,
"learning_rate": 8.276551724137932e-06,
"loss": 0.2952,
"step": 3000
},
{
"epoch": 3.04,
"eval_loss": 0.3534850776195526,
"eval_runtime": 37910.4054,
"eval_samples_per_second": 1.361,
"eval_steps_per_second": 0.17,
"eval_wer": 18.584123758335593,
"step": 3000
},
{
"epoch": 3.05,
"learning_rate": 8.207586206896553e-06,
"loss": 0.2521,
"step": 3100
},
{
"epoch": 3.05,
"learning_rate": 8.138620689655173e-06,
"loss": 0.2683,
"step": 3200
},
{
"epoch": 4.01,
"learning_rate": 8.069655172413794e-06,
"loss": 0.204,
"step": 3300
},
{
"epoch": 4.01,
"learning_rate": 8.000689655172415e-06,
"loss": 0.3076,
"step": 3400
},
{
"epoch": 4.02,
"learning_rate": 7.931724137931034e-06,
"loss": 0.2536,
"step": 3500
},
{
"epoch": 4.03,
"learning_rate": 7.862758620689656e-06,
"loss": 0.1466,
"step": 3600
},
{
"epoch": 4.03,
"learning_rate": 7.793793103448277e-06,
"loss": 0.2349,
"step": 3700
},
{
"epoch": 4.04,
"learning_rate": 7.724827586206896e-06,
"loss": 0.0717,
"step": 3800
},
{
"epoch": 4.05,
"learning_rate": 7.655862068965517e-06,
"loss": 0.1646,
"step": 3900
},
{
"epoch": 4.05,
"learning_rate": 7.586896551724139e-06,
"loss": 0.1876,
"step": 4000
},
{
"epoch": 4.05,
"eval_loss": 0.34643468260765076,
"eval_runtime": 37497.7523,
"eval_samples_per_second": 1.376,
"eval_steps_per_second": 0.172,
"eval_wer": 17.50968055751862,
"step": 4000
},
{
"epoch": 5.0,
"learning_rate": 7.517931034482759e-06,
"loss": 0.11,
"step": 4100
},
{
"epoch": 5.01,
"learning_rate": 7.44896551724138e-06,
"loss": 0.1532,
"step": 4200
},
{
"epoch": 5.02,
"learning_rate": 7.3800000000000005e-06,
"loss": 0.1261,
"step": 4300
},
{
"epoch": 5.03,
"learning_rate": 7.311724137931035e-06,
"loss": 0.3031,
"step": 4400
},
{
"epoch": 5.03,
"learning_rate": 7.242758620689656e-06,
"loss": 0.2008,
"step": 4500
},
{
"epoch": 5.04,
"learning_rate": 7.173793103448277e-06,
"loss": 0.1994,
"step": 4600
},
{
"epoch": 5.04,
"learning_rate": 7.104827586206897e-06,
"loss": 0.1358,
"step": 4700
},
{
"epoch": 5.05,
"learning_rate": 7.0358620689655175e-06,
"loss": 0.1003,
"step": 4800
},
{
"epoch": 6.0,
"learning_rate": 6.966896551724139e-06,
"loss": 0.0803,
"step": 4900
},
{
"epoch": 6.01,
"learning_rate": 6.897931034482759e-06,
"loss": 0.1037,
"step": 5000
},
{
"epoch": 6.01,
"eval_loss": 0.3396442234516144,
"eval_runtime": 37513.698,
"eval_samples_per_second": 1.376,
"eval_steps_per_second": 0.172,
"eval_wer": 16.73599995925316,
"step": 5000
},
{
"epoch": 6.02,
"learning_rate": 6.829655172413794e-06,
"loss": 0.0979,
"step": 5100
},
{
"epoch": 6.02,
"learning_rate": 6.760689655172414e-06,
"loss": 0.1106,
"step": 5200
},
{
"epoch": 6.03,
"learning_rate": 6.691724137931035e-06,
"loss": 0.0544,
"step": 5300
},
{
"epoch": 6.04,
"learning_rate": 6.622758620689656e-06,
"loss": 0.108,
"step": 5400
},
{
"epoch": 6.04,
"learning_rate": 6.553793103448276e-06,
"loss": 0.1078,
"step": 5500
},
{
"epoch": 6.05,
"learning_rate": 6.484827586206896e-06,
"loss": 0.0944,
"step": 5600
},
{
"epoch": 7.0,
"learning_rate": 6.415862068965518e-06,
"loss": 0.0823,
"step": 5700
},
{
"epoch": 7.01,
"learning_rate": 6.346896551724139e-06,
"loss": 0.1471,
"step": 5800
},
{
"epoch": 7.02,
"learning_rate": 6.277931034482759e-06,
"loss": 0.1238,
"step": 5900
},
{
"epoch": 7.02,
"learning_rate": 6.20896551724138e-06,
"loss": 0.0473,
"step": 6000
},
{
"epoch": 7.02,
"eval_loss": 0.35263511538505554,
"eval_runtime": 37260.0087,
"eval_samples_per_second": 1.385,
"eval_steps_per_second": 0.173,
"eval_wer": 16.41308126320293,
"step": 6000
},
{
"epoch": 7.03,
"learning_rate": 6.1400000000000005e-06,
"loss": 0.1612,
"step": 6100
},
{
"epoch": 7.04,
"learning_rate": 6.071034482758621e-06,
"loss": 0.1428,
"step": 6200
},
{
"epoch": 7.04,
"learning_rate": 6.002068965517241e-06,
"loss": 0.0976,
"step": 6300
},
{
"epoch": 7.05,
"learning_rate": 5.933103448275862e-06,
"loss": 0.0992,
"step": 6400
},
{
"epoch": 8.0,
"learning_rate": 5.8641379310344835e-06,
"loss": 0.151,
"step": 6500
},
{
"epoch": 8.01,
"learning_rate": 5.795172413793104e-06,
"loss": 0.1433,
"step": 6600
},
{
"epoch": 8.02,
"learning_rate": 5.726206896551725e-06,
"loss": 0.0398,
"step": 6700
},
{
"epoch": 8.02,
"learning_rate": 5.657241379310345e-06,
"loss": 0.0515,
"step": 6800
},
{
"epoch": 8.03,
"learning_rate": 5.588275862068966e-06,
"loss": 0.0578,
"step": 6900
},
{
"epoch": 8.04,
"learning_rate": 5.519310344827587e-06,
"loss": 0.1605,
"step": 7000
},
{
"epoch": 8.04,
"eval_loss": 0.3284248411655426,
"eval_runtime": 36674.0508,
"eval_samples_per_second": 1.407,
"eval_steps_per_second": 0.176,
"eval_wer": 16.40123921322948,
"step": 7000
},
{
"epoch": 8.04,
"learning_rate": 5.450344827586207e-06,
"loss": 0.1565,
"step": 7100
},
{
"epoch": 8.05,
"learning_rate": 5.3813793103448275e-06,
"loss": 0.1178,
"step": 7200
},
{
"epoch": 9.0,
"learning_rate": 5.3124137931034495e-06,
"loss": 0.1594,
"step": 7300
},
{
"epoch": 9.01,
"learning_rate": 5.24344827586207e-06,
"loss": 0.1737,
"step": 7400
},
{
"epoch": 9.02,
"learning_rate": 5.17448275862069e-06,
"loss": 0.0702,
"step": 7500
},
{
"epoch": 9.02,
"learning_rate": 5.1055172413793105e-06,
"loss": 0.1071,
"step": 7600
},
{
"epoch": 9.03,
"learning_rate": 5.036551724137932e-06,
"loss": 0.0482,
"step": 7700
},
{
"epoch": 9.04,
"learning_rate": 4.967586206896552e-06,
"loss": 0.0774,
"step": 7800
},
{
"epoch": 9.04,
"learning_rate": 4.898620689655173e-06,
"loss": 0.1031,
"step": 7900
},
{
"epoch": 9.05,
"learning_rate": 4.8296551724137935e-06,
"loss": 0.0537,
"step": 8000
},
{
"epoch": 9.05,
"eval_loss": 0.33857810497283936,
"eval_runtime": 36709.3165,
"eval_samples_per_second": 1.406,
"eval_steps_per_second": 0.176,
"eval_wer": 15.945383956186962,
"step": 8000
},
{
"epoch": 10.0,
"learning_rate": 4.760689655172414e-06,
"loss": 0.0315,
"step": 8100
},
{
"epoch": 10.01,
"learning_rate": 4.691724137931035e-06,
"loss": 0.0335,
"step": 8200
},
{
"epoch": 10.02,
"learning_rate": 4.622758620689655e-06,
"loss": 0.0826,
"step": 8300
},
{
"epoch": 10.02,
"learning_rate": 4.5537931034482765e-06,
"loss": 0.0876,
"step": 8400
},
{
"epoch": 10.03,
"learning_rate": 4.484827586206897e-06,
"loss": 0.0548,
"step": 8500
},
{
"epoch": 10.04,
"learning_rate": 4.415862068965517e-06,
"loss": 0.086,
"step": 8600
},
{
"epoch": 10.04,
"learning_rate": 4.346896551724138e-06,
"loss": 0.1344,
"step": 8700
},
{
"epoch": 10.05,
"learning_rate": 4.277931034482759e-06,
"loss": 0.1348,
"step": 8800
},
{
"epoch": 11.0,
"learning_rate": 4.20896551724138e-06,
"loss": 0.1021,
"step": 8900
},
{
"epoch": 11.01,
"learning_rate": 4.14e-06,
"loss": 0.0928,
"step": 9000
},
{
"epoch": 11.01,
"eval_loss": 0.3314945697784424,
"eval_runtime": 37649.5122,
"eval_samples_per_second": 1.371,
"eval_steps_per_second": 0.171,
"eval_wer": 15.956844004548365,
"step": 9000
},
{
"epoch": 11.02,
"learning_rate": 4.071034482758621e-06,
"loss": 0.1342,
"step": 9100
},
{
"epoch": 11.02,
"learning_rate": 4.002068965517242e-06,
"loss": 0.1719,
"step": 9200
},
{
"epoch": 11.03,
"learning_rate": 3.933103448275862e-06,
"loss": 0.1673,
"step": 9300
},
{
"epoch": 11.04,
"learning_rate": 3.864137931034483e-06,
"loss": 0.1906,
"step": 9400
},
{
"epoch": 11.04,
"learning_rate": 3.795172413793104e-06,
"loss": 0.1577,
"step": 9500
},
{
"epoch": 11.05,
"learning_rate": 3.7262068965517247e-06,
"loss": 0.0188,
"step": 9600
},
{
"epoch": 12.0,
"learning_rate": 3.657241379310345e-06,
"loss": 0.0797,
"step": 9700
},
{
"epoch": 12.01,
"learning_rate": 3.5882758620689658e-06,
"loss": 0.0657,
"step": 9800
},
{
"epoch": 12.02,
"learning_rate": 3.5193103448275865e-06,
"loss": 0.0996,
"step": 9900
},
{
"epoch": 12.02,
"learning_rate": 3.4503448275862073e-06,
"loss": 0.0144,
"step": 10000
},
{
"epoch": 12.02,
"eval_loss": 0.35319629311561584,
"eval_runtime": 37622.7796,
"eval_samples_per_second": 1.372,
"eval_steps_per_second": 0.171,
"eval_wer": 15.538679573227798,
"step": 10000
},
{
"epoch": 12.03,
"learning_rate": 3.381379310344828e-06,
"loss": 0.0491,
"step": 10100
},
{
"epoch": 12.04,
"learning_rate": 3.3124137931034483e-06,
"loss": 0.0418,
"step": 10200
},
{
"epoch": 12.04,
"learning_rate": 3.2434482758620695e-06,
"loss": 0.0454,
"step": 10300
},
{
"epoch": 12.05,
"learning_rate": 3.17448275862069e-06,
"loss": 0.0162,
"step": 10400
},
{
"epoch": 13.0,
"learning_rate": 3.1055172413793106e-06,
"loss": 0.0325,
"step": 10500
},
{
"epoch": 13.01,
"learning_rate": 3.036551724137931e-06,
"loss": 0.0746,
"step": 10600
},
{
"epoch": 13.02,
"learning_rate": 2.967586206896552e-06,
"loss": 0.2385,
"step": 10700
},
{
"epoch": 13.02,
"learning_rate": 2.898620689655173e-06,
"loss": 0.2072,
"step": 10800
},
{
"epoch": 13.03,
"learning_rate": 2.829655172413793e-06,
"loss": 0.0993,
"step": 10900
},
{
"epoch": 13.04,
"learning_rate": 2.760689655172414e-06,
"loss": 0.0267,
"step": 11000
},
{
"epoch": 13.04,
"eval_loss": 0.32611754536628723,
"eval_runtime": 37216.3199,
"eval_samples_per_second": 1.387,
"eval_steps_per_second": 0.173,
"eval_wer": 15.757693830801301,
"step": 11000
},
{
"epoch": 13.04,
"learning_rate": 2.691724137931035e-06,
"loss": 0.0645,
"step": 11100
},
{
"epoch": 13.05,
"learning_rate": 2.6227586206896554e-06,
"loss": 0.0366,
"step": 11200
},
{
"epoch": 14.0,
"learning_rate": 2.553793103448276e-06,
"loss": 0.0959,
"step": 11300
},
{
"epoch": 14.01,
"learning_rate": 2.484827586206897e-06,
"loss": 0.046,
"step": 11400
},
{
"epoch": 14.02,
"learning_rate": 2.4165517241379314e-06,
"loss": 0.0808,
"step": 11500
},
{
"epoch": 14.02,
"learning_rate": 2.3475862068965517e-06,
"loss": 0.0356,
"step": 11600
},
{
"epoch": 14.03,
"learning_rate": 2.278620689655173e-06,
"loss": 0.0455,
"step": 11700
},
{
"epoch": 14.04,
"learning_rate": 2.209655172413793e-06,
"loss": 0.1122,
"step": 11800
},
{
"epoch": 14.04,
"learning_rate": 2.140689655172414e-06,
"loss": 0.1622,
"step": 11900
},
{
"epoch": 14.05,
"learning_rate": 2.0717241379310347e-06,
"loss": 0.0936,
"step": 12000
},
{
"epoch": 14.05,
"eval_loss": 0.3155108094215393,
"eval_runtime": 36574.1071,
"eval_samples_per_second": 1.411,
"eval_steps_per_second": 0.176,
"eval_wer": 15.338001393032547,
"step": 12000
},
{
"epoch": 15.0,
"learning_rate": 2.0027586206896554e-06,
"loss": 0.0741,
"step": 12100
},
{
"epoch": 15.01,
"learning_rate": 1.933793103448276e-06,
"loss": 0.118,
"step": 12200
},
{
"epoch": 15.02,
"learning_rate": 1.8648275862068967e-06,
"loss": 0.0423,
"step": 12300
},
{
"epoch": 15.02,
"learning_rate": 1.7958620689655173e-06,
"loss": 0.0356,
"step": 12400
},
{
"epoch": 15.03,
"learning_rate": 1.7268965517241382e-06,
"loss": 0.0316,
"step": 12500
},
{
"epoch": 15.04,
"learning_rate": 1.6579310344827588e-06,
"loss": 0.0473,
"step": 12600
},
{
"epoch": 15.04,
"learning_rate": 1.5889655172413795e-06,
"loss": 0.0775,
"step": 12700
},
{
"epoch": 15.05,
"learning_rate": 1.52e-06,
"loss": 0.0709,
"step": 12800
},
{
"epoch": 16.0,
"learning_rate": 1.4510344827586208e-06,
"loss": 0.1303,
"step": 12900
},
{
"epoch": 16.01,
"learning_rate": 1.3820689655172416e-06,
"loss": 0.0825,
"step": 13000
},
{
"epoch": 16.01,
"eval_loss": 0.31983643770217896,
"eval_runtime": 36792.1562,
"eval_samples_per_second": 1.403,
"eval_steps_per_second": 0.175,
"eval_wer": 15.26529375287297,
"step": 13000
},
{
"epoch": 16.01,
"learning_rate": 1.3131034482758623e-06,
"loss": 0.0235,
"step": 13100
},
{
"epoch": 16.02,
"learning_rate": 1.2441379310344829e-06,
"loss": 0.0359,
"step": 13200
},
{
"epoch": 16.03,
"learning_rate": 1.1751724137931036e-06,
"loss": 0.0375,
"step": 13300
},
{
"epoch": 16.03,
"learning_rate": 1.1062068965517241e-06,
"loss": 0.0338,
"step": 13400
},
{
"epoch": 16.04,
"learning_rate": 1.0372413793103449e-06,
"loss": 0.0264,
"step": 13500
},
{
"epoch": 16.05,
"learning_rate": 9.682758620689656e-07,
"loss": 0.0199,
"step": 13600
},
{
"epoch": 17.0,
"learning_rate": 8.993103448275863e-07,
"loss": 0.027,
"step": 13700
},
{
"epoch": 17.01,
"learning_rate": 8.310344827586207e-07,
"loss": 0.0367,
"step": 13800
},
{
"epoch": 17.01,
"learning_rate": 7.620689655172415e-07,
"loss": 0.0895,
"step": 13900
},
{
"epoch": 17.02,
"learning_rate": 6.931034482758621e-07,
"loss": 0.0498,
"step": 14000
},
{
"epoch": 17.02,
"eval_loss": 0.3385982811450958,
"eval_runtime": 37095.9082,
"eval_samples_per_second": 1.391,
"eval_steps_per_second": 0.174,
"eval_wer": 15.197170132057957,
"step": 14000
},
{
"epoch": 17.03,
"learning_rate": 6.241379310344828e-07,
"loss": 0.0778,
"step": 14100
},
{
"epoch": 17.03,
"learning_rate": 5.551724137931035e-07,
"loss": 0.098,
"step": 14200
},
{
"epoch": 17.04,
"learning_rate": 4.862068965517241e-07,
"loss": 0.0203,
"step": 14300
},
{
"epoch": 17.05,
"learning_rate": 4.1724137931034485e-07,
"loss": 0.0328,
"step": 14400
},
{
"epoch": 18.0,
"learning_rate": 3.4827586206896555e-07,
"loss": 0.0132,
"step": 14500
},
{
"epoch": 18.01,
"learning_rate": 2.7931034482758624e-07,
"loss": 0.0195,
"step": 14600
},
{
"epoch": 18.01,
"learning_rate": 2.1034482758620692e-07,
"loss": 0.0244,
"step": 14700
},
{
"epoch": 18.02,
"learning_rate": 1.413793103448276e-07,
"loss": 0.0382,
"step": 14800
},
{
"epoch": 18.03,
"learning_rate": 7.241379310344829e-08,
"loss": 0.0354,
"step": 14900
},
{
"epoch": 18.03,
"learning_rate": 3.4482758620689654e-09,
"loss": 0.0338,
"step": 15000
},
{
"epoch": 18.03,
"eval_loss": 0.341310977935791,
"eval_runtime": 40247.8087,
"eval_samples_per_second": 1.282,
"eval_steps_per_second": 0.16,
"eval_wer": 15.197170132057957,
"step": 15000
}
],
"max_steps": 15000,
"num_train_epochs": 9223372036854775807,
"total_flos": 2.77157418467328e+20,
"trial_name": null,
"trial_params": null
}