1011 lines
23 KiB
JSON
1011 lines
23 KiB
JSON
|
{
|
||
|
|
"best_metric": null,
|
||
|
|
"best_model_checkpoint": null,
|
||
|
|
"epoch": 35.9993690851735,
|
||
|
|
"global_step": 28512,
|
||
|
|
"is_hyper_param_search": false,
|
||
|
|
"is_local_process_zero": true,
|
||
|
|
"is_world_process_zero": true,
|
||
|
|
"log_history": [
|
||
|
|
{
|
||
|
|
"epoch": 0.5,
|
||
|
|
"learning_rate": 6.400000000000001e-05,
|
||
|
|
"loss": 12.3462,
|
||
|
|
"step": 400
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 0.5,
|
||
|
|
"eval_cer": 0.8798367057152999,
|
||
|
|
"eval_loss": 5.513787269592285,
|
||
|
|
"eval_runtime": 24.8146,
|
||
|
|
"eval_samples_per_second": 20.835,
|
||
|
|
"step": 400
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 1.01,
|
||
|
|
"learning_rate": 7.938618925831203e-05,
|
||
|
|
"loss": 4.5486,
|
||
|
|
"step": 800
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 1.01,
|
||
|
|
"eval_cer": 0.6050763223287184,
|
||
|
|
"eval_loss": 3.230196237564087,
|
||
|
|
"eval_runtime": 23.8368,
|
||
|
|
"eval_samples_per_second": 21.689,
|
||
|
|
"step": 800
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 1.51,
|
||
|
|
"learning_rate": 7.856777493606139e-05,
|
||
|
|
"loss": 2.7575,
|
||
|
|
"step": 1200
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 1.51,
|
||
|
|
"eval_cer": 0.4703585374511892,
|
||
|
|
"eval_loss": 2.1326839923858643,
|
||
|
|
"eval_runtime": 23.7759,
|
||
|
|
"eval_samples_per_second": 21.745,
|
||
|
|
"step": 1200
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 2.02,
|
||
|
|
"learning_rate": 7.774936061381075e-05,
|
||
|
|
"loss": 1.943,
|
||
|
|
"step": 1600
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 2.02,
|
||
|
|
"eval_cer": 0.39882854100106496,
|
||
|
|
"eval_loss": 1.5994508266448975,
|
||
|
|
"eval_runtime": 21.9909,
|
||
|
|
"eval_samples_per_second": 23.51,
|
||
|
|
"step": 1600
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 2.52,
|
||
|
|
"learning_rate": 7.693094629156011e-05,
|
||
|
|
"loss": 1.4609,
|
||
|
|
"step": 2000
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 2.52,
|
||
|
|
"eval_cer": 0.3842740504082357,
|
||
|
|
"eval_loss": 1.4217835664749146,
|
||
|
|
"eval_runtime": 21.7005,
|
||
|
|
"eval_samples_per_second": 23.824,
|
||
|
|
"step": 2000
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 3.03,
|
||
|
|
"learning_rate": 7.611253196930947e-05,
|
||
|
|
"loss": 1.2765,
|
||
|
|
"step": 2400
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 3.03,
|
||
|
|
"eval_cer": 0.36705715299964503,
|
||
|
|
"eval_loss": 1.2836037874221802,
|
||
|
|
"eval_runtime": 22.9952,
|
||
|
|
"eval_samples_per_second": 22.483,
|
||
|
|
"step": 2400
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 3.54,
|
||
|
|
"learning_rate": 7.529411764705883e-05,
|
||
|
|
"loss": 1.0495,
|
||
|
|
"step": 2800
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 3.54,
|
||
|
|
"eval_cer": 0.36421725239616615,
|
||
|
|
"eval_loss": 1.2246943712234497,
|
||
|
|
"eval_runtime": 22.6153,
|
||
|
|
"eval_samples_per_second": 22.861,
|
||
|
|
"step": 2800
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 4.04,
|
||
|
|
"learning_rate": 7.447570332480819e-05,
|
||
|
|
"loss": 0.9616,
|
||
|
|
"step": 3200
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 4.04,
|
||
|
|
"eval_cer": 0.34185303514376997,
|
||
|
|
"eval_loss": 1.1541343927383423,
|
||
|
|
"eval_runtime": 22.0945,
|
||
|
|
"eval_samples_per_second": 23.399,
|
||
|
|
"step": 3200
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 4.55,
|
||
|
|
"learning_rate": 7.365728900255755e-05,
|
||
|
|
"loss": 0.8222,
|
||
|
|
"step": 3600
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 4.55,
|
||
|
|
"eval_cer": 0.3331558395456159,
|
||
|
|
"eval_loss": 1.1167793273925781,
|
||
|
|
"eval_runtime": 22.9218,
|
||
|
|
"eval_samples_per_second": 22.555,
|
||
|
|
"step": 3600
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 5.05,
|
||
|
|
"learning_rate": 7.283887468030691e-05,
|
||
|
|
"loss": 0.789,
|
||
|
|
"step": 4000
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 5.05,
|
||
|
|
"eval_cer": 0.3347532836350728,
|
||
|
|
"eval_loss": 1.0833882093429565,
|
||
|
|
"eval_runtime": 22.1912,
|
||
|
|
"eval_samples_per_second": 23.298,
|
||
|
|
"step": 4000
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 5.56,
|
||
|
|
"learning_rate": 7.202046035805627e-05,
|
||
|
|
"loss": 0.6838,
|
||
|
|
"step": 4400
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 5.56,
|
||
|
|
"eval_cer": 0.3196663116790912,
|
||
|
|
"eval_loss": 1.0477311611175537,
|
||
|
|
"eval_runtime": 21.6837,
|
||
|
|
"eval_samples_per_second": 23.843,
|
||
|
|
"step": 4400
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 6.06,
|
||
|
|
"learning_rate": 7.120204603580563e-05,
|
||
|
|
"loss": 0.6487,
|
||
|
|
"step": 4800
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 6.06,
|
||
|
|
"eval_cer": 0.3184238551650692,
|
||
|
|
"eval_loss": 1.036700963973999,
|
||
|
|
"eval_runtime": 22.8347,
|
||
|
|
"eval_samples_per_second": 22.641,
|
||
|
|
"step": 4800
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 6.57,
|
||
|
|
"learning_rate": 7.0383631713555e-05,
|
||
|
|
"loss": 0.5654,
|
||
|
|
"step": 5200
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 6.57,
|
||
|
|
"eval_cer": 0.31576144834930775,
|
||
|
|
"eval_loss": 1.017685890197754,
|
||
|
|
"eval_runtime": 21.8232,
|
||
|
|
"eval_samples_per_second": 23.69,
|
||
|
|
"step": 5200
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 7.07,
|
||
|
|
"learning_rate": 6.956521739130436e-05,
|
||
|
|
"loss": 0.5529,
|
||
|
|
"step": 5600
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 7.07,
|
||
|
|
"eval_cer": 0.3106141285055023,
|
||
|
|
"eval_loss": 1.010923981666565,
|
||
|
|
"eval_runtime": 22.2843,
|
||
|
|
"eval_samples_per_second": 23.2,
|
||
|
|
"step": 5600
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 7.58,
|
||
|
|
"learning_rate": 6.874680306905372e-05,
|
||
|
|
"loss": 0.4907,
|
||
|
|
"step": 6000
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 7.58,
|
||
|
|
"eval_cer": 0.3061767838125666,
|
||
|
|
"eval_loss": 1.0107142925262451,
|
||
|
|
"eval_runtime": 21.6939,
|
||
|
|
"eval_samples_per_second": 23.832,
|
||
|
|
"step": 6000
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 8.08,
|
||
|
|
"learning_rate": 6.792838874680308e-05,
|
||
|
|
"loss": 0.4712,
|
||
|
|
"step": 6400
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 8.08,
|
||
|
|
"eval_cer": 0.3065317713880014,
|
||
|
|
"eval_loss": 1.0030597448349,
|
||
|
|
"eval_runtime": 22.8516,
|
||
|
|
"eval_samples_per_second": 22.624,
|
||
|
|
"step": 6400
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 8.59,
|
||
|
|
"learning_rate": 6.710997442455244e-05,
|
||
|
|
"loss": 0.4183,
|
||
|
|
"step": 6800
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 8.59,
|
||
|
|
"eval_cer": 0.30564430244941426,
|
||
|
|
"eval_loss": 0.9820496439933777,
|
||
|
|
"eval_runtime": 21.5904,
|
||
|
|
"eval_samples_per_second": 23.946,
|
||
|
|
"step": 6800
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 9.09,
|
||
|
|
"learning_rate": 6.62915601023018e-05,
|
||
|
|
"loss": 0.4065,
|
||
|
|
"step": 7200
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 9.09,
|
||
|
|
"eval_cer": 0.30138445154419596,
|
||
|
|
"eval_loss": 0.9998520612716675,
|
||
|
|
"eval_runtime": 22.1151,
|
||
|
|
"eval_samples_per_second": 23.378,
|
||
|
|
"step": 7200
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 9.6,
|
||
|
|
"learning_rate": 6.547314578005116e-05,
|
||
|
|
"loss": 0.3744,
|
||
|
|
"step": 7600
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 9.6,
|
||
|
|
"eval_cer": 0.29907703230386934,
|
||
|
|
"eval_loss": 1.000787377357483,
|
||
|
|
"eval_runtime": 22.3274,
|
||
|
|
"eval_samples_per_second": 23.155,
|
||
|
|
"step": 7600
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 10.1,
|
||
|
|
"learning_rate": 6.465473145780052e-05,
|
||
|
|
"loss": 0.3592,
|
||
|
|
"step": 8000
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 10.1,
|
||
|
|
"eval_cer": 0.2946396876109336,
|
||
|
|
"eval_loss": 0.9794895052909851,
|
||
|
|
"eval_runtime": 22.123,
|
||
|
|
"eval_samples_per_second": 23.369,
|
||
|
|
"step": 8000
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 10.61,
|
||
|
|
"learning_rate": 6.383631713554988e-05,
|
||
|
|
"loss": 0.322,
|
||
|
|
"step": 8400
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 10.61,
|
||
|
|
"eval_cer": 0.29818956336528224,
|
||
|
|
"eval_loss": 1.0048284530639648,
|
||
|
|
"eval_runtime": 22.2712,
|
||
|
|
"eval_samples_per_second": 23.214,
|
||
|
|
"step": 8400
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 11.11,
|
||
|
|
"learning_rate": 6.301790281329924e-05,
|
||
|
|
"loss": 0.3362,
|
||
|
|
"step": 8800
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 11.11,
|
||
|
|
"eval_cer": 0.2942847000354988,
|
||
|
|
"eval_loss": 0.9930339455604553,
|
||
|
|
"eval_runtime": 21.9289,
|
||
|
|
"eval_samples_per_second": 23.576,
|
||
|
|
"step": 8800
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 11.62,
|
||
|
|
"learning_rate": 6.219948849104859e-05,
|
||
|
|
"loss": 0.3006,
|
||
|
|
"step": 9200
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 11.62,
|
||
|
|
"eval_cer": 0.29499467518636846,
|
||
|
|
"eval_loss": 0.9977458119392395,
|
||
|
|
"eval_runtime": 21.4941,
|
||
|
|
"eval_samples_per_second": 24.053,
|
||
|
|
"step": 9200
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 12.12,
|
||
|
|
"learning_rate": 6.138107416879796e-05,
|
||
|
|
"loss": 0.2836,
|
||
|
|
"step": 9600
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 12.12,
|
||
|
|
"eval_cer": 0.28895988640397585,
|
||
|
|
"eval_loss": 0.9851016998291016,
|
||
|
|
"eval_runtime": 22.2205,
|
||
|
|
"eval_samples_per_second": 23.267,
|
||
|
|
"step": 9600
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 12.63,
|
||
|
|
"learning_rate": 6.056265984654732e-05,
|
||
|
|
"loss": 0.2674,
|
||
|
|
"step": 10000
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 12.63,
|
||
|
|
"eval_cer": 0.28647497337593186,
|
||
|
|
"eval_loss": 0.984660804271698,
|
||
|
|
"eval_runtime": 22.17,
|
||
|
|
"eval_samples_per_second": 23.32,
|
||
|
|
"step": 10000
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 13.13,
|
||
|
|
"learning_rate": 5.974424552429668e-05,
|
||
|
|
"loss": 0.2584,
|
||
|
|
"step": 10400
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 13.13,
|
||
|
|
"eval_cer": 0.28611998580049697,
|
||
|
|
"eval_loss": 0.9858261942863464,
|
||
|
|
"eval_runtime": 21.856,
|
||
|
|
"eval_samples_per_second": 23.655,
|
||
|
|
"step": 10400
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 13.64,
|
||
|
|
"learning_rate": 5.8925831202046044e-05,
|
||
|
|
"loss": 0.2386,
|
||
|
|
"step": 10800
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 13.64,
|
||
|
|
"eval_cer": 0.2854100106496273,
|
||
|
|
"eval_loss": 0.9940375685691833,
|
||
|
|
"eval_runtime": 22.1157,
|
||
|
|
"eval_samples_per_second": 23.377,
|
||
|
|
"step": 10800
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 14.14,
|
||
|
|
"learning_rate": 5.81074168797954e-05,
|
||
|
|
"loss": 0.2322,
|
||
|
|
"step": 11200
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 14.14,
|
||
|
|
"eval_cer": 0.2884274050408236,
|
||
|
|
"eval_loss": 1.0060980319976807,
|
||
|
|
"eval_runtime": 22.0771,
|
||
|
|
"eval_samples_per_second": 23.418,
|
||
|
|
"step": 11200
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 14.65,
|
||
|
|
"learning_rate": 5.728900255754476e-05,
|
||
|
|
"loss": 0.2093,
|
||
|
|
"step": 11600
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 14.65,
|
||
|
|
"eval_cer": 0.2870074547390841,
|
||
|
|
"eval_loss": 1.0193942785263062,
|
||
|
|
"eval_runtime": 22.6977,
|
||
|
|
"eval_samples_per_second": 22.778,
|
||
|
|
"step": 11600
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 15.15,
|
||
|
|
"learning_rate": 5.6470588235294126e-05,
|
||
|
|
"loss": 0.2103,
|
||
|
|
"step": 12000
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 15.15,
|
||
|
|
"eval_cer": 0.2854100106496273,
|
||
|
|
"eval_loss": 1.0035730600357056,
|
||
|
|
"eval_runtime": 22.0407,
|
||
|
|
"eval_samples_per_second": 23.457,
|
||
|
|
"step": 12000
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 15.66,
|
||
|
|
"learning_rate": 5.565217391304348e-05,
|
||
|
|
"loss": 0.197,
|
||
|
|
"step": 12400
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 15.66,
|
||
|
|
"eval_cer": 0.2820376286829961,
|
||
|
|
"eval_loss": 1.0057300329208374,
|
||
|
|
"eval_runtime": 22.058,
|
||
|
|
"eval_samples_per_second": 23.438,
|
||
|
|
"step": 12400
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 16.16,
|
||
|
|
"learning_rate": 5.483375959079285e-05,
|
||
|
|
"loss": 0.1965,
|
||
|
|
"step": 12800
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 16.16,
|
||
|
|
"eval_cer": 0.28345757898473556,
|
||
|
|
"eval_loss": 1.0166577100753784,
|
||
|
|
"eval_runtime": 22.8481,
|
||
|
|
"eval_samples_per_second": 22.628,
|
||
|
|
"step": 12800
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 16.67,
|
||
|
|
"learning_rate": 5.40153452685422e-05,
|
||
|
|
"loss": 0.1867,
|
||
|
|
"step": 13200
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 16.67,
|
||
|
|
"eval_cer": 0.2784877529286475,
|
||
|
|
"eval_loss": 1.017500400543213,
|
||
|
|
"eval_runtime": 22.4539,
|
||
|
|
"eval_samples_per_second": 23.025,
|
||
|
|
"step": 13200
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 17.17,
|
||
|
|
"learning_rate": 5.319693094629156e-05,
|
||
|
|
"loss": 0.1766,
|
||
|
|
"step": 13600
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 17.17,
|
||
|
|
"eval_cer": 0.2797302094426695,
|
||
|
|
"eval_loss": 1.0199682712554932,
|
||
|
|
"eval_runtime": 22.2314,
|
||
|
|
"eval_samples_per_second": 23.255,
|
||
|
|
"step": 13600
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 17.68,
|
||
|
|
"learning_rate": 5.237851662404093e-05,
|
||
|
|
"loss": 0.1744,
|
||
|
|
"step": 14000
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 17.68,
|
||
|
|
"eval_cer": 0.28186013489527867,
|
||
|
|
"eval_loss": 1.0218894481658936,
|
||
|
|
"eval_runtime": 22.2184,
|
||
|
|
"eval_samples_per_second": 23.269,
|
||
|
|
"step": 14000
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 18.18,
|
||
|
|
"learning_rate": 5.1560102301790284e-05,
|
||
|
|
"loss": 0.1651,
|
||
|
|
"step": 14400
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 18.18,
|
||
|
|
"eval_cer": 0.2825701100461484,
|
||
|
|
"eval_loss": 1.0378596782684326,
|
||
|
|
"eval_runtime": 24.2738,
|
||
|
|
"eval_samples_per_second": 21.299,
|
||
|
|
"step": 14400
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 18.69,
|
||
|
|
"learning_rate": 5.074168797953965e-05,
|
||
|
|
"loss": 0.1584,
|
||
|
|
"step": 14800
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 18.69,
|
||
|
|
"eval_cer": 0.2790202342917998,
|
||
|
|
"eval_loss": 1.0371792316436768,
|
||
|
|
"eval_runtime": 23.2497,
|
||
|
|
"eval_samples_per_second": 22.237,
|
||
|
|
"step": 14800
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 19.19,
|
||
|
|
"learning_rate": 4.9923273657289005e-05,
|
||
|
|
"loss": 0.1569,
|
||
|
|
"step": 15200
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 19.19,
|
||
|
|
"eval_cer": 0.2793752218672346,
|
||
|
|
"eval_loss": 1.030590534210205,
|
||
|
|
"eval_runtime": 23.4394,
|
||
|
|
"eval_samples_per_second": 22.057,
|
||
|
|
"step": 15200
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 19.7,
|
||
|
|
"learning_rate": 4.9104859335038366e-05,
|
||
|
|
"loss": 0.1435,
|
||
|
|
"step": 15600
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 19.7,
|
||
|
|
"eval_cer": 0.28115015974440893,
|
||
|
|
"eval_loss": 1.0243384838104248,
|
||
|
|
"eval_runtime": 23.5764,
|
||
|
|
"eval_samples_per_second": 21.929,
|
||
|
|
"step": 15600
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 20.2,
|
||
|
|
"learning_rate": 4.8286445012787726e-05,
|
||
|
|
"loss": 0.1494,
|
||
|
|
"step": 16000
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 20.2,
|
||
|
|
"eval_cer": 0.28434504792332266,
|
||
|
|
"eval_loss": 1.0423582792282104,
|
||
|
|
"eval_runtime": 21.5787,
|
||
|
|
"eval_samples_per_second": 23.959,
|
||
|
|
"step": 16000
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 20.71,
|
||
|
|
"learning_rate": 4.746803069053709e-05,
|
||
|
|
"loss": 0.1458,
|
||
|
|
"step": 16400
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 20.71,
|
||
|
|
"eval_cer": 0.27706780262690806,
|
||
|
|
"eval_loss": 1.014714002609253,
|
||
|
|
"eval_runtime": 22.6407,
|
||
|
|
"eval_samples_per_second": 22.835,
|
||
|
|
"step": 16400
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 21.21,
|
||
|
|
"learning_rate": 4.6649616368286454e-05,
|
||
|
|
"loss": 0.1304,
|
||
|
|
"step": 16800
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 21.21,
|
||
|
|
"eval_cer": 0.2809726659566915,
|
||
|
|
"eval_loss": 1.0456069707870483,
|
||
|
|
"eval_runtime": 22.1413,
|
||
|
|
"eval_samples_per_second": 23.35,
|
||
|
|
"step": 16800
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 21.72,
|
||
|
|
"learning_rate": 4.583120204603581e-05,
|
||
|
|
"loss": 0.1309,
|
||
|
|
"step": 17200
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 21.72,
|
||
|
|
"eval_cer": 0.2760028399006035,
|
||
|
|
"eval_loss": 1.0279470682144165,
|
||
|
|
"eval_runtime": 22.2014,
|
||
|
|
"eval_samples_per_second": 23.287,
|
||
|
|
"step": 17200
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 22.22,
|
||
|
|
"learning_rate": 4.501278772378517e-05,
|
||
|
|
"loss": 0.1241,
|
||
|
|
"step": 17600
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 22.22,
|
||
|
|
"eval_cer": 0.2747603833865815,
|
||
|
|
"eval_loss": 1.0260205268859863,
|
||
|
|
"eval_runtime": 22.1272,
|
||
|
|
"eval_samples_per_second": 23.365,
|
||
|
|
"step": 17600
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 22.73,
|
||
|
|
"learning_rate": 4.419437340153453e-05,
|
||
|
|
"loss": 0.1209,
|
||
|
|
"step": 18000
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 22.73,
|
||
|
|
"eval_cer": 0.27547035853745117,
|
||
|
|
"eval_loss": 1.0386370420455933,
|
||
|
|
"eval_runtime": 22.1033,
|
||
|
|
"eval_samples_per_second": 23.39,
|
||
|
|
"step": 18000
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 23.23,
|
||
|
|
"learning_rate": 4.337595907928389e-05,
|
||
|
|
"loss": 0.1182,
|
||
|
|
"step": 18400
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 23.23,
|
||
|
|
"eval_cer": 0.2731629392971246,
|
||
|
|
"eval_loss": 1.0359835624694824,
|
||
|
|
"eval_runtime": 22.1075,
|
||
|
|
"eval_samples_per_second": 23.386,
|
||
|
|
"step": 18400
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 23.74,
|
||
|
|
"learning_rate": 4.255754475703326e-05,
|
||
|
|
"loss": 0.1119,
|
||
|
|
"step": 18800
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 23.74,
|
||
|
|
"eval_cer": 0.2751153709620163,
|
||
|
|
"eval_loss": 1.0298691987991333,
|
||
|
|
"eval_runtime": 23.4709,
|
||
|
|
"eval_samples_per_second": 22.027,
|
||
|
|
"step": 18800
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 24.24,
|
||
|
|
"learning_rate": 4.173913043478261e-05,
|
||
|
|
"loss": 0.1291,
|
||
|
|
"step": 19200
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 24.24,
|
||
|
|
"eval_cer": 0.27547035853745117,
|
||
|
|
"eval_loss": 1.0163068771362305,
|
||
|
|
"eval_runtime": 24.0816,
|
||
|
|
"eval_samples_per_second": 21.469,
|
||
|
|
"step": 19200
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 24.75,
|
||
|
|
"learning_rate": 4.092071611253197e-05,
|
||
|
|
"loss": 0.1478,
|
||
|
|
"step": 19600
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 24.75,
|
||
|
|
"eval_cer": 0.2729854455094072,
|
||
|
|
"eval_loss": 1.0182946920394897,
|
||
|
|
"eval_runtime": 23.6272,
|
||
|
|
"eval_samples_per_second": 21.882,
|
||
|
|
"step": 19600
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 25.25,
|
||
|
|
"learning_rate": 4.010230179028133e-05,
|
||
|
|
"loss": 0.149,
|
||
|
|
"step": 20000
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 25.25,
|
||
|
|
"eval_cer": 0.27689030883919064,
|
||
|
|
"eval_loss": 1.0228219032287598,
|
||
|
|
"eval_runtime": 22.6194,
|
||
|
|
"eval_samples_per_second": 22.857,
|
||
|
|
"step": 20000
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 25.76,
|
||
|
|
"learning_rate": 3.9283887468030694e-05,
|
||
|
|
"loss": 0.1517,
|
||
|
|
"step": 20400
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 25.76,
|
||
|
|
"eval_cer": 0.2777777777777778,
|
||
|
|
"eval_loss": 1.013156771659851,
|
||
|
|
"eval_runtime": 22.4049,
|
||
|
|
"eval_samples_per_second": 23.075,
|
||
|
|
"step": 20400
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 26.26,
|
||
|
|
"learning_rate": 3.8465473145780055e-05,
|
||
|
|
"loss": 0.1427,
|
||
|
|
"step": 20800
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 26.26,
|
||
|
|
"eval_cer": 0.27405040823571175,
|
||
|
|
"eval_loss": 1.0105297565460205,
|
||
|
|
"eval_runtime": 23.1206,
|
||
|
|
"eval_samples_per_second": 22.361,
|
||
|
|
"step": 20800
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 26.77,
|
||
|
|
"learning_rate": 3.7647058823529415e-05,
|
||
|
|
"loss": 0.142,
|
||
|
|
"step": 21200
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 26.77,
|
||
|
|
"eval_cer": 0.27422790202342917,
|
||
|
|
"eval_loss": 1.0065898895263672,
|
||
|
|
"eval_runtime": 23.5837,
|
||
|
|
"eval_samples_per_second": 21.922,
|
||
|
|
"step": 21200
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 27.27,
|
||
|
|
"learning_rate": 3.6828644501278776e-05,
|
||
|
|
"loss": 0.146,
|
||
|
|
"step": 21600
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 27.27,
|
||
|
|
"eval_cer": 0.27582534611288606,
|
||
|
|
"eval_loss": 1.0007686614990234,
|
||
|
|
"eval_runtime": 22.8462,
|
||
|
|
"eval_samples_per_second": 22.63,
|
||
|
|
"step": 21600
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 27.78,
|
||
|
|
"learning_rate": 3.601023017902814e-05,
|
||
|
|
"loss": 0.1395,
|
||
|
|
"step": 22000
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 27.78,
|
||
|
|
"eval_cer": 0.2731629392971246,
|
||
|
|
"eval_loss": 1.0027023553848267,
|
||
|
|
"eval_runtime": 22.6777,
|
||
|
|
"eval_samples_per_second": 22.798,
|
||
|
|
"step": 22000
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 28.28,
|
||
|
|
"learning_rate": 3.51918158567775e-05,
|
||
|
|
"loss": 0.1326,
|
||
|
|
"step": 22400
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 28.28,
|
||
|
|
"eval_cer": 0.2763578274760383,
|
||
|
|
"eval_loss": 1.0187015533447266,
|
||
|
|
"eval_runtime": 24.8634,
|
||
|
|
"eval_samples_per_second": 20.794,
|
||
|
|
"step": 22400
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 28.79,
|
||
|
|
"learning_rate": 3.437340153452686e-05,
|
||
|
|
"loss": 0.1352,
|
||
|
|
"step": 22800
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 28.79,
|
||
|
|
"eval_cer": 0.2713880014199503,
|
||
|
|
"eval_loss": 1.0136111974716187,
|
||
|
|
"eval_runtime": 24.4177,
|
||
|
|
"eval_samples_per_second": 21.173,
|
||
|
|
"step": 22800
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 29.29,
|
||
|
|
"learning_rate": 3.355498721227622e-05,
|
||
|
|
"loss": 0.1285,
|
||
|
|
"step": 23200
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 29.29,
|
||
|
|
"eval_cer": 0.27689030883919064,
|
||
|
|
"eval_loss": 1.018797755241394,
|
||
|
|
"eval_runtime": 22.5421,
|
||
|
|
"eval_samples_per_second": 22.935,
|
||
|
|
"step": 23200
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 29.8,
|
||
|
|
"learning_rate": 3.273657289002558e-05,
|
||
|
|
"loss": 0.1309,
|
||
|
|
"step": 23600
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 29.8,
|
||
|
|
"eval_cer": 0.273340433084842,
|
||
|
|
"eval_loss": 1.0131142139434814,
|
||
|
|
"eval_runtime": 26.263,
|
||
|
|
"eval_samples_per_second": 19.686,
|
||
|
|
"step": 23600
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 30.3,
|
||
|
|
"learning_rate": 3.191815856777494e-05,
|
||
|
|
"loss": 0.1213,
|
||
|
|
"step": 24000
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 30.3,
|
||
|
|
"eval_cer": 0.27050053248136313,
|
||
|
|
"eval_loss": 1.0099353790283203,
|
||
|
|
"eval_runtime": 26.863,
|
||
|
|
"eval_samples_per_second": 19.246,
|
||
|
|
"step": 24000
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 30.81,
|
||
|
|
"learning_rate": 3.1099744245524294e-05,
|
||
|
|
"loss": 0.1294,
|
||
|
|
"step": 24400
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 30.81,
|
||
|
|
"eval_cer": 0.26943556975505856,
|
||
|
|
"eval_loss": 1.000162124633789,
|
||
|
|
"eval_runtime": 25.0465,
|
||
|
|
"eval_samples_per_second": 20.642,
|
||
|
|
"step": 24400
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 31.31,
|
||
|
|
"learning_rate": 3.028132992327366e-05,
|
||
|
|
"loss": 0.1222,
|
||
|
|
"step": 24800
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 31.31,
|
||
|
|
"eval_cer": 0.26819311324103656,
|
||
|
|
"eval_loss": 0.9948088526725769,
|
||
|
|
"eval_runtime": 24.7877,
|
||
|
|
"eval_samples_per_second": 20.857,
|
||
|
|
"step": 24800
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 31.82,
|
||
|
|
"learning_rate": 2.9462915601023022e-05,
|
||
|
|
"loss": 0.1277,
|
||
|
|
"step": 25200
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 31.82,
|
||
|
|
"eval_cer": 0.26943556975505856,
|
||
|
|
"eval_loss": 0.9982617497444153,
|
||
|
|
"eval_runtime": 26.2619,
|
||
|
|
"eval_samples_per_second": 19.686,
|
||
|
|
"step": 25200
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 32.32,
|
||
|
|
"learning_rate": 2.864450127877238e-05,
|
||
|
|
"loss": 0.1195,
|
||
|
|
"step": 25600
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 32.32,
|
||
|
|
"eval_cer": 0.26979055733049345,
|
||
|
|
"eval_loss": 0.9970006346702576,
|
||
|
|
"eval_runtime": 21.895,
|
||
|
|
"eval_samples_per_second": 23.613,
|
||
|
|
"step": 25600
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 32.83,
|
||
|
|
"learning_rate": 2.782608695652174e-05,
|
||
|
|
"loss": 0.115,
|
||
|
|
"step": 26000
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 32.83,
|
||
|
|
"eval_cer": 0.2729854455094072,
|
||
|
|
"eval_loss": 0.9976910352706909,
|
||
|
|
"eval_runtime": 21.8104,
|
||
|
|
"eval_samples_per_second": 23.704,
|
||
|
|
"step": 26000
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 33.33,
|
||
|
|
"learning_rate": 2.70076726342711e-05,
|
||
|
|
"loss": 0.1239,
|
||
|
|
"step": 26400
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 33.33,
|
||
|
|
"eval_cer": 0.2689030883919063,
|
||
|
|
"eval_loss": 0.9979565739631653,
|
||
|
|
"eval_runtime": 22.8985,
|
||
|
|
"eval_samples_per_second": 22.578,
|
||
|
|
"step": 26400
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 33.84,
|
||
|
|
"learning_rate": 2.6189258312020465e-05,
|
||
|
|
"loss": 0.1191,
|
||
|
|
"step": 26800
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 33.84,
|
||
|
|
"eval_cer": 0.2674831380901668,
|
||
|
|
"eval_loss": 1.0030490159988403,
|
||
|
|
"eval_runtime": 28.0306,
|
||
|
|
"eval_samples_per_second": 18.444,
|
||
|
|
"step": 26800
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 34.34,
|
||
|
|
"learning_rate": 2.5370843989769826e-05,
|
||
|
|
"loss": 0.1125,
|
||
|
|
"step": 27200
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 34.34,
|
||
|
|
"eval_cer": 0.269613063542776,
|
||
|
|
"eval_loss": 1.0033153295516968,
|
||
|
|
"eval_runtime": 26.4843,
|
||
|
|
"eval_samples_per_second": 19.521,
|
||
|
|
"step": 27200
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 34.85,
|
||
|
|
"learning_rate": 2.4552429667519183e-05,
|
||
|
|
"loss": 0.1126,
|
||
|
|
"step": 27600
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 34.85,
|
||
|
|
"eval_cer": 0.26695065672701457,
|
||
|
|
"eval_loss": 0.997735321521759,
|
||
|
|
"eval_runtime": 26.1441,
|
||
|
|
"eval_samples_per_second": 19.775,
|
||
|
|
"step": 27600
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 35.35,
|
||
|
|
"learning_rate": 2.3734015345268544e-05,
|
||
|
|
"loss": 0.1158,
|
||
|
|
"step": 28000
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 35.35,
|
||
|
|
"eval_cer": 0.26996805111821087,
|
||
|
|
"eval_loss": 1.0106626749038696,
|
||
|
|
"eval_runtime": 21.2071,
|
||
|
|
"eval_samples_per_second": 24.379,
|
||
|
|
"step": 28000
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 35.86,
|
||
|
|
"learning_rate": 2.2915601023017904e-05,
|
||
|
|
"loss": 0.1184,
|
||
|
|
"step": 28400
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 35.86,
|
||
|
|
"eval_cer": 0.26819311324103656,
|
||
|
|
"eval_loss": 1.005273699760437,
|
||
|
|
"eval_runtime": 25.2781,
|
||
|
|
"eval_samples_per_second": 20.453,
|
||
|
|
"step": 28400
|
||
|
|
}
|
||
|
|
],
|
||
|
|
"max_steps": 39600,
|
||
|
|
"num_train_epochs": 50,
|
||
|
|
"total_flos": 5.74598534526558e+19,
|
||
|
|
"trial_name": null,
|
||
|
|
"trial_params": null
|
||
|
|
}
|