LexLLMv0.0.0.x.10.9d / trainer_state.json
andrealexroom's picture
Upload folder using huggingface_hub
394b1c9 verified
{
"best_metric": null,
"best_model_checkpoint": null,
"epoch": 1.3183926616980246,
"eval_steps": 93,
"global_step": 1581,
"is_hyper_param_search": false,
"is_local_process_zero": true,
"is_world_process_zero": true,
"log_history": [
{
"epoch": 0.0,
"learning_rate": 1e-05,
"loss": 22.2771,
"step": 1
},
{
"epoch": 0.08,
"learning_rate": 1e-05,
"loss": 3.3736,
"step": 93
},
{
"epoch": 0.08,
"eval_validation_privacy_sources_loss": 1.4779064655303955,
"eval_validation_privacy_sources_runtime": 169.7235,
"eval_validation_privacy_sources_samples_per_second": 1.355,
"eval_validation_privacy_sources_steps_per_second": 0.678,
"step": 93
},
{
"epoch": 0.08,
"eval_validation_agenda_digitale_loss": 2.152458429336548,
"eval_validation_agenda_digitale_runtime": 160.5968,
"eval_validation_agenda_digitale_samples_per_second": 1.357,
"eval_validation_agenda_digitale_steps_per_second": 0.679,
"step": 93
},
{
"epoch": 0.08,
"eval_validation_leggepertutti_loss": 1.8981273174285889,
"eval_validation_leggepertutti_runtime": 28.0044,
"eval_validation_leggepertutti_samples_per_second": 1.357,
"eval_validation_leggepertutti_steps_per_second": 0.678,
"step": 93
},
{
"epoch": 0.08,
"eval_validation_altalexprivacy_loss": 1.9700795412063599,
"eval_validation_altalexprivacy_runtime": 45.7,
"eval_validation_altalexprivacy_samples_per_second": 1.357,
"eval_validation_altalexprivacy_steps_per_second": 0.678,
"step": 93
},
{
"epoch": 0.16,
"learning_rate": 1e-05,
"loss": 1.7135,
"step": 186
},
{
"epoch": 0.16,
"eval_validation_privacy_sources_loss": 1.289480209350586,
"eval_validation_privacy_sources_runtime": 169.4222,
"eval_validation_privacy_sources_samples_per_second": 1.358,
"eval_validation_privacy_sources_steps_per_second": 0.679,
"step": 186
},
{
"epoch": 0.16,
"eval_validation_agenda_digitale_loss": 1.9645200967788696,
"eval_validation_agenda_digitale_runtime": 160.5935,
"eval_validation_agenda_digitale_samples_per_second": 1.357,
"eval_validation_agenda_digitale_steps_per_second": 0.679,
"step": 186
},
{
"epoch": 0.16,
"eval_validation_leggepertutti_loss": 1.7435646057128906,
"eval_validation_leggepertutti_runtime": 27.9989,
"eval_validation_leggepertutti_samples_per_second": 1.357,
"eval_validation_leggepertutti_steps_per_second": 0.679,
"step": 186
},
{
"epoch": 0.16,
"eval_validation_altalexprivacy_loss": 1.7977099418640137,
"eval_validation_altalexprivacy_runtime": 45.6746,
"eval_validation_altalexprivacy_samples_per_second": 1.357,
"eval_validation_altalexprivacy_steps_per_second": 0.679,
"step": 186
},
{
"epoch": 0.23,
"learning_rate": 1e-05,
"loss": 1.5958,
"step": 279
},
{
"epoch": 0.23,
"eval_validation_privacy_sources_loss": 1.2112126350402832,
"eval_validation_privacy_sources_runtime": 169.4752,
"eval_validation_privacy_sources_samples_per_second": 1.357,
"eval_validation_privacy_sources_steps_per_second": 0.679,
"step": 279
},
{
"epoch": 0.23,
"eval_validation_agenda_digitale_loss": 1.8859468698501587,
"eval_validation_agenda_digitale_runtime": 160.6253,
"eval_validation_agenda_digitale_samples_per_second": 1.357,
"eval_validation_agenda_digitale_steps_per_second": 0.679,
"step": 279
},
{
"epoch": 0.23,
"eval_validation_leggepertutti_loss": 1.6768569946289062,
"eval_validation_leggepertutti_runtime": 28.0039,
"eval_validation_leggepertutti_samples_per_second": 1.357,
"eval_validation_leggepertutti_steps_per_second": 0.678,
"step": 279
},
{
"epoch": 0.23,
"eval_validation_altalexprivacy_loss": 1.7135707139968872,
"eval_validation_altalexprivacy_runtime": 45.6744,
"eval_validation_altalexprivacy_samples_per_second": 1.357,
"eval_validation_altalexprivacy_steps_per_second": 0.679,
"step": 279
},
{
"epoch": 0.31,
"learning_rate": 1e-05,
"loss": 1.5249,
"step": 372
},
{
"epoch": 0.31,
"eval_validation_privacy_sources_loss": 1.1636546850204468,
"eval_validation_privacy_sources_runtime": 169.3483,
"eval_validation_privacy_sources_samples_per_second": 1.358,
"eval_validation_privacy_sources_steps_per_second": 0.679,
"step": 372
},
{
"epoch": 0.31,
"eval_validation_agenda_digitale_loss": 1.839213490486145,
"eval_validation_agenda_digitale_runtime": 160.5345,
"eval_validation_agenda_digitale_samples_per_second": 1.358,
"eval_validation_agenda_digitale_steps_per_second": 0.679,
"step": 372
},
{
"epoch": 0.31,
"eval_validation_leggepertutti_loss": 1.6400021314620972,
"eval_validation_leggepertutti_runtime": 27.9724,
"eval_validation_leggepertutti_samples_per_second": 1.358,
"eval_validation_leggepertutti_steps_per_second": 0.679,
"step": 372
},
{
"epoch": 0.31,
"eval_validation_altalexprivacy_loss": 1.6672061681747437,
"eval_validation_altalexprivacy_runtime": 45.6542,
"eval_validation_altalexprivacy_samples_per_second": 1.358,
"eval_validation_altalexprivacy_steps_per_second": 0.679,
"step": 372
},
{
"epoch": 0.39,
"learning_rate": 1e-05,
"loss": 1.4843,
"step": 465
},
{
"epoch": 0.39,
"eval_validation_privacy_sources_loss": 1.1293103694915771,
"eval_validation_privacy_sources_runtime": 169.8454,
"eval_validation_privacy_sources_samples_per_second": 1.354,
"eval_validation_privacy_sources_steps_per_second": 0.677,
"step": 465
},
{
"epoch": 0.39,
"eval_validation_agenda_digitale_loss": 1.8078429698944092,
"eval_validation_agenda_digitale_runtime": 160.9923,
"eval_validation_agenda_digitale_samples_per_second": 1.354,
"eval_validation_agenda_digitale_steps_per_second": 0.677,
"step": 465
},
{
"epoch": 0.39,
"eval_validation_leggepertutti_loss": 1.6059364080429077,
"eval_validation_leggepertutti_runtime": 28.0726,
"eval_validation_leggepertutti_samples_per_second": 1.354,
"eval_validation_leggepertutti_steps_per_second": 0.677,
"step": 465
},
{
"epoch": 0.39,
"eval_validation_altalexprivacy_loss": 1.62712562084198,
"eval_validation_altalexprivacy_runtime": 45.8103,
"eval_validation_altalexprivacy_samples_per_second": 1.353,
"eval_validation_altalexprivacy_steps_per_second": 0.677,
"step": 465
},
{
"epoch": 0.47,
"learning_rate": 1e-05,
"loss": 1.4525,
"step": 558
},
{
"epoch": 0.47,
"eval_validation_privacy_sources_loss": 1.1031160354614258,
"eval_validation_privacy_sources_runtime": 169.8338,
"eval_validation_privacy_sources_samples_per_second": 1.354,
"eval_validation_privacy_sources_steps_per_second": 0.677,
"step": 558
},
{
"epoch": 0.47,
"eval_validation_agenda_digitale_loss": 1.7819572687149048,
"eval_validation_agenda_digitale_runtime": 160.9065,
"eval_validation_agenda_digitale_samples_per_second": 1.355,
"eval_validation_agenda_digitale_steps_per_second": 0.677,
"step": 558
},
{
"epoch": 0.47,
"eval_validation_leggepertutti_loss": 1.585649847984314,
"eval_validation_leggepertutti_runtime": 28.0472,
"eval_validation_leggepertutti_samples_per_second": 1.355,
"eval_validation_leggepertutti_steps_per_second": 0.677,
"step": 558
},
{
"epoch": 0.47,
"eval_validation_altalexprivacy_loss": 1.6030628681182861,
"eval_validation_altalexprivacy_runtime": 45.7744,
"eval_validation_altalexprivacy_samples_per_second": 1.354,
"eval_validation_altalexprivacy_steps_per_second": 0.677,
"step": 558
},
{
"epoch": 0.54,
"learning_rate": 1e-05,
"loss": 1.4045,
"step": 651
},
{
"epoch": 0.54,
"eval_validation_privacy_sources_loss": 1.076947808265686,
"eval_validation_privacy_sources_runtime": 169.8094,
"eval_validation_privacy_sources_samples_per_second": 1.354,
"eval_validation_privacy_sources_steps_per_second": 0.677,
"step": 651
},
{
"epoch": 0.54,
"eval_validation_agenda_digitale_loss": 1.7600467205047607,
"eval_validation_agenda_digitale_runtime": 160.9231,
"eval_validation_agenda_digitale_samples_per_second": 1.355,
"eval_validation_agenda_digitale_steps_per_second": 0.677,
"step": 651
},
{
"epoch": 0.54,
"eval_validation_leggepertutti_loss": 1.568356990814209,
"eval_validation_leggepertutti_runtime": 28.0579,
"eval_validation_leggepertutti_samples_per_second": 1.354,
"eval_validation_leggepertutti_steps_per_second": 0.677,
"step": 651
},
{
"epoch": 0.54,
"eval_validation_altalexprivacy_loss": 1.5807151794433594,
"eval_validation_altalexprivacy_runtime": 45.7761,
"eval_validation_altalexprivacy_samples_per_second": 1.354,
"eval_validation_altalexprivacy_steps_per_second": 0.677,
"step": 651
},
{
"epoch": 0.62,
"learning_rate": 1e-05,
"loss": 1.3958,
"step": 744
},
{
"epoch": 0.62,
"eval_validation_privacy_sources_loss": 1.0633326768875122,
"eval_validation_privacy_sources_runtime": 169.7153,
"eval_validation_privacy_sources_samples_per_second": 1.355,
"eval_validation_privacy_sources_steps_per_second": 0.678,
"step": 744
},
{
"epoch": 0.62,
"eval_validation_agenda_digitale_loss": 1.7451695203781128,
"eval_validation_agenda_digitale_runtime": 160.896,
"eval_validation_agenda_digitale_samples_per_second": 1.355,
"eval_validation_agenda_digitale_steps_per_second": 0.677,
"step": 744
},
{
"epoch": 0.62,
"eval_validation_leggepertutti_loss": 1.5583051443099976,
"eval_validation_leggepertutti_runtime": 28.048,
"eval_validation_leggepertutti_samples_per_second": 1.355,
"eval_validation_leggepertutti_steps_per_second": 0.677,
"step": 744
},
{
"epoch": 0.62,
"eval_validation_altalexprivacy_loss": 1.5672190189361572,
"eval_validation_altalexprivacy_runtime": 45.7525,
"eval_validation_altalexprivacy_samples_per_second": 1.355,
"eval_validation_altalexprivacy_steps_per_second": 0.678,
"step": 744
},
{
"epoch": 0.7,
"learning_rate": 1e-05,
"loss": 1.3715,
"step": 837
},
{
"epoch": 0.7,
"eval_validation_privacy_sources_loss": 1.046081304550171,
"eval_validation_privacy_sources_runtime": 169.714,
"eval_validation_privacy_sources_samples_per_second": 1.355,
"eval_validation_privacy_sources_steps_per_second": 0.678,
"step": 837
},
{
"epoch": 0.7,
"eval_validation_agenda_digitale_loss": 1.7291052341461182,
"eval_validation_agenda_digitale_runtime": 160.8608,
"eval_validation_agenda_digitale_samples_per_second": 1.355,
"eval_validation_agenda_digitale_steps_per_second": 0.678,
"step": 837
},
{
"epoch": 0.7,
"eval_validation_leggepertutti_loss": 1.5531929731369019,
"eval_validation_leggepertutti_runtime": 28.0451,
"eval_validation_leggepertutti_samples_per_second": 1.355,
"eval_validation_leggepertutti_steps_per_second": 0.677,
"step": 837
},
{
"epoch": 0.7,
"eval_validation_altalexprivacy_loss": 1.5526759624481201,
"eval_validation_altalexprivacy_runtime": 45.7626,
"eval_validation_altalexprivacy_samples_per_second": 1.355,
"eval_validation_altalexprivacy_steps_per_second": 0.677,
"step": 837
},
{
"epoch": 0.78,
"learning_rate": 1e-05,
"loss": 1.3555,
"step": 930
},
{
"epoch": 0.78,
"eval_validation_privacy_sources_loss": 1.0322965383529663,
"eval_validation_privacy_sources_runtime": 169.6103,
"eval_validation_privacy_sources_samples_per_second": 1.356,
"eval_validation_privacy_sources_steps_per_second": 0.678,
"step": 930
},
{
"epoch": 0.78,
"eval_validation_agenda_digitale_loss": 1.7139462232589722,
"eval_validation_agenda_digitale_runtime": 160.7355,
"eval_validation_agenda_digitale_samples_per_second": 1.356,
"eval_validation_agenda_digitale_steps_per_second": 0.678,
"step": 930
},
{
"epoch": 0.78,
"eval_validation_leggepertutti_loss": 1.5367027521133423,
"eval_validation_leggepertutti_runtime": 28.027,
"eval_validation_leggepertutti_samples_per_second": 1.356,
"eval_validation_leggepertutti_steps_per_second": 0.678,
"step": 930
},
{
"epoch": 0.78,
"eval_validation_altalexprivacy_loss": 1.5342811346054077,
"eval_validation_altalexprivacy_runtime": 45.7283,
"eval_validation_altalexprivacy_samples_per_second": 1.356,
"eval_validation_altalexprivacy_steps_per_second": 0.678,
"step": 930
},
{
"epoch": 0.85,
"learning_rate": 1e-05,
"loss": 1.3566,
"step": 1023
},
{
"epoch": 0.85,
"eval_validation_privacy_sources_loss": 1.022092342376709,
"eval_validation_privacy_sources_runtime": 169.4239,
"eval_validation_privacy_sources_samples_per_second": 1.358,
"eval_validation_privacy_sources_steps_per_second": 0.679,
"step": 1023
},
{
"epoch": 0.85,
"eval_validation_agenda_digitale_loss": 1.7054636478424072,
"eval_validation_agenda_digitale_runtime": 160.5379,
"eval_validation_agenda_digitale_samples_per_second": 1.358,
"eval_validation_agenda_digitale_steps_per_second": 0.679,
"step": 1023
},
{
"epoch": 0.85,
"eval_validation_leggepertutti_loss": 1.5282613039016724,
"eval_validation_leggepertutti_runtime": 27.9854,
"eval_validation_leggepertutti_samples_per_second": 1.358,
"eval_validation_leggepertutti_steps_per_second": 0.679,
"step": 1023
},
{
"epoch": 0.85,
"eval_validation_altalexprivacy_loss": 1.5194894075393677,
"eval_validation_altalexprivacy_runtime": 45.6648,
"eval_validation_altalexprivacy_samples_per_second": 1.358,
"eval_validation_altalexprivacy_steps_per_second": 0.679,
"step": 1023
},
{
"epoch": 0.93,
"learning_rate": 1e-05,
"loss": 1.3279,
"step": 1116
},
{
"epoch": 0.93,
"eval_validation_privacy_sources_loss": 1.0133976936340332,
"eval_validation_privacy_sources_runtime": 169.5192,
"eval_validation_privacy_sources_samples_per_second": 1.357,
"eval_validation_privacy_sources_steps_per_second": 0.678,
"step": 1116
},
{
"epoch": 0.93,
"eval_validation_agenda_digitale_loss": 1.6934592723846436,
"eval_validation_agenda_digitale_runtime": 160.703,
"eval_validation_agenda_digitale_samples_per_second": 1.357,
"eval_validation_agenda_digitale_steps_per_second": 0.678,
"step": 1116
},
{
"epoch": 0.93,
"eval_validation_leggepertutti_loss": 1.5167473554611206,
"eval_validation_leggepertutti_runtime": 28.0019,
"eval_validation_leggepertutti_samples_per_second": 1.357,
"eval_validation_leggepertutti_steps_per_second": 0.679,
"step": 1116
},
{
"epoch": 0.93,
"eval_validation_altalexprivacy_loss": 1.5086991786956787,
"eval_validation_altalexprivacy_runtime": 45.6936,
"eval_validation_altalexprivacy_samples_per_second": 1.357,
"eval_validation_altalexprivacy_steps_per_second": 0.678,
"step": 1116
},
{
"epoch": 1.01,
"learning_rate": 4.000000000000001e-06,
"loss": 1.32,
"step": 1209
},
{
"epoch": 1.01,
"eval_validation_privacy_sources_loss": 1.0032093524932861,
"eval_validation_privacy_sources_runtime": 169.5456,
"eval_validation_privacy_sources_samples_per_second": 1.357,
"eval_validation_privacy_sources_steps_per_second": 0.678,
"step": 1209
},
{
"epoch": 1.01,
"eval_validation_agenda_digitale_loss": 1.68604576587677,
"eval_validation_agenda_digitale_runtime": 160.6963,
"eval_validation_agenda_digitale_samples_per_second": 1.357,
"eval_validation_agenda_digitale_steps_per_second": 0.678,
"step": 1209
},
{
"epoch": 1.01,
"eval_validation_leggepertutti_loss": 1.5239654779434204,
"eval_validation_leggepertutti_runtime": 28.0205,
"eval_validation_leggepertutti_samples_per_second": 1.356,
"eval_validation_leggepertutti_steps_per_second": 0.678,
"step": 1209
},
{
"epoch": 1.01,
"eval_validation_altalexprivacy_loss": 1.5007380247116089,
"eval_validation_altalexprivacy_runtime": 45.7063,
"eval_validation_altalexprivacy_samples_per_second": 1.356,
"eval_validation_altalexprivacy_steps_per_second": 0.678,
"step": 1209
},
{
"epoch": 1.09,
"learning_rate": 4.000000000000001e-06,
"loss": 1.1998,
"step": 1302
},
{
"epoch": 1.09,
"eval_validation_privacy_sources_loss": 0.9907514452934265,
"eval_validation_privacy_sources_runtime": 169.4228,
"eval_validation_privacy_sources_samples_per_second": 1.358,
"eval_validation_privacy_sources_steps_per_second": 0.679,
"step": 1302
},
{
"epoch": 1.09,
"eval_validation_agenda_digitale_loss": 1.6758071184158325,
"eval_validation_agenda_digitale_runtime": 160.6151,
"eval_validation_agenda_digitale_samples_per_second": 1.357,
"eval_validation_agenda_digitale_steps_per_second": 0.679,
"step": 1302
},
{
"epoch": 1.09,
"eval_validation_leggepertutti_loss": 1.5173066854476929,
"eval_validation_leggepertutti_runtime": 27.9869,
"eval_validation_leggepertutti_samples_per_second": 1.358,
"eval_validation_leggepertutti_steps_per_second": 0.679,
"step": 1302
},
{
"epoch": 1.09,
"eval_validation_altalexprivacy_loss": 1.4873181581497192,
"eval_validation_altalexprivacy_runtime": 45.6834,
"eval_validation_altalexprivacy_samples_per_second": 1.357,
"eval_validation_altalexprivacy_steps_per_second": 0.679,
"step": 1302
},
{
"epoch": 1.16,
"learning_rate": 4.000000000000001e-06,
"loss": 1.1635,
"step": 1395
},
{
"epoch": 1.16,
"eval_validation_privacy_sources_loss": 0.98543781042099,
"eval_validation_privacy_sources_runtime": 169.7781,
"eval_validation_privacy_sources_samples_per_second": 1.355,
"eval_validation_privacy_sources_steps_per_second": 0.677,
"step": 1395
},
{
"epoch": 1.16,
"eval_validation_agenda_digitale_loss": 1.671740174293518,
"eval_validation_agenda_digitale_runtime": 160.8659,
"eval_validation_agenda_digitale_samples_per_second": 1.355,
"eval_validation_agenda_digitale_steps_per_second": 0.678,
"step": 1395
},
{
"epoch": 1.16,
"eval_validation_leggepertutti_loss": 1.5131033658981323,
"eval_validation_leggepertutti_runtime": 28.0459,
"eval_validation_leggepertutti_samples_per_second": 1.355,
"eval_validation_leggepertutti_steps_per_second": 0.677,
"step": 1395
},
{
"epoch": 1.16,
"eval_validation_altalexprivacy_loss": 1.4810400009155273,
"eval_validation_altalexprivacy_runtime": 45.7413,
"eval_validation_altalexprivacy_samples_per_second": 1.355,
"eval_validation_altalexprivacy_steps_per_second": 0.678,
"step": 1395
},
{
"epoch": 1.24,
"learning_rate": 4.000000000000001e-06,
"loss": 1.188,
"step": 1488
},
{
"epoch": 1.24,
"eval_validation_privacy_sources_loss": 0.981711745262146,
"eval_validation_privacy_sources_runtime": 169.7712,
"eval_validation_privacy_sources_samples_per_second": 1.355,
"eval_validation_privacy_sources_steps_per_second": 0.677,
"step": 1488
},
{
"epoch": 1.24,
"eval_validation_agenda_digitale_loss": 1.669861078262329,
"eval_validation_agenda_digitale_runtime": 160.8963,
"eval_validation_agenda_digitale_samples_per_second": 1.355,
"eval_validation_agenda_digitale_steps_per_second": 0.677,
"step": 1488
},
{
"epoch": 1.24,
"eval_validation_leggepertutti_loss": 1.5106905698776245,
"eval_validation_leggepertutti_runtime": 28.0461,
"eval_validation_leggepertutti_samples_per_second": 1.355,
"eval_validation_leggepertutti_steps_per_second": 0.677,
"step": 1488
},
{
"epoch": 1.24,
"eval_validation_altalexprivacy_loss": 1.4784834384918213,
"eval_validation_altalexprivacy_runtime": 45.7714,
"eval_validation_altalexprivacy_samples_per_second": 1.355,
"eval_validation_altalexprivacy_steps_per_second": 0.677,
"step": 1488
},
{
"epoch": 1.32,
"learning_rate": 4.000000000000001e-06,
"loss": 1.169,
"step": 1581
},
{
"epoch": 1.32,
"eval_validation_privacy_sources_loss": 0.9772719740867615,
"eval_validation_privacy_sources_runtime": 169.5316,
"eval_validation_privacy_sources_samples_per_second": 1.357,
"eval_validation_privacy_sources_steps_per_second": 0.678,
"step": 1581
},
{
"epoch": 1.32,
"eval_validation_agenda_digitale_loss": 1.6676079034805298,
"eval_validation_agenda_digitale_runtime": 160.8887,
"eval_validation_agenda_digitale_samples_per_second": 1.355,
"eval_validation_agenda_digitale_steps_per_second": 0.677,
"step": 1581
},
{
"epoch": 1.32,
"eval_validation_leggepertutti_loss": 1.5096626281738281,
"eval_validation_leggepertutti_runtime": 28.0405,
"eval_validation_leggepertutti_samples_per_second": 1.355,
"eval_validation_leggepertutti_steps_per_second": 0.678,
"step": 1581
},
{
"epoch": 1.32,
"eval_validation_altalexprivacy_loss": 1.4747135639190674,
"eval_validation_altalexprivacy_runtime": 45.7877,
"eval_validation_altalexprivacy_samples_per_second": 1.354,
"eval_validation_altalexprivacy_steps_per_second": 0.677,
"step": 1581
}
],
"logging_steps": 93,
"max_steps": 2398,
"num_input_tokens_seen": 0,
"num_train_epochs": 2,
"save_steps": 93,
"total_flos": 1.7741598460471149e+19,
"train_batch_size": 1,
"trial_name": null,
"trial_params": null
}