LexLLMv0.0.0.x.10.14c / trainer_state.json
andrealexroom's picture
Upload folder using huggingface_hub
eff98e9 verified
{
"best_metric": null,
"best_model_checkpoint": null,
"epoch": 0.385245839722623,
"eval_steps": 375,
"global_step": 6375,
"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": 1.8182,
"step": 1
},
{
"epoch": 0.02,
"learning_rate": 1e-05,
"loss": 1.8531,
"step": 375
},
{
"epoch": 0.02,
"eval_validation_privacy_sources_loss": 1.0427056550979614,
"eval_validation_privacy_sources_runtime": 201.1093,
"eval_validation_privacy_sources_samples_per_second": 1.904,
"eval_validation_privacy_sources_steps_per_second": 1.904,
"step": 375
},
{
"epoch": 0.02,
"eval_validation_agenda_digitale_loss": 1.5768671035766602,
"eval_validation_agenda_digitale_runtime": 178.6225,
"eval_validation_agenda_digitale_samples_per_second": 1.903,
"eval_validation_agenda_digitale_steps_per_second": 1.903,
"step": 375
},
{
"epoch": 0.02,
"eval_validation_leggepertutti_loss": 1.41618013381958,
"eval_validation_leggepertutti_runtime": 34.1702,
"eval_validation_leggepertutti_samples_per_second": 1.902,
"eval_validation_leggepertutti_steps_per_second": 1.902,
"step": 375
},
{
"epoch": 0.02,
"eval_validation_altalexprivacy_loss": 1.420487880706787,
"eval_validation_altalexprivacy_runtime": 55.6679,
"eval_validation_altalexprivacy_samples_per_second": 1.904,
"eval_validation_altalexprivacy_steps_per_second": 1.904,
"step": 375
},
{
"epoch": 0.02,
"eval_validation_mc4_it_loss": 2.0156803131103516,
"eval_validation_mc4_it_runtime": 74.5861,
"eval_validation_mc4_it_samples_per_second": 1.904,
"eval_validation_mc4_it_steps_per_second": 1.904,
"step": 375
},
{
"epoch": 0.05,
"learning_rate": 1e-05,
"loss": 1.659,
"step": 750
},
{
"epoch": 0.05,
"eval_validation_privacy_sources_loss": 0.9875672459602356,
"eval_validation_privacy_sources_runtime": 201.0647,
"eval_validation_privacy_sources_samples_per_second": 1.905,
"eval_validation_privacy_sources_steps_per_second": 1.905,
"step": 750
},
{
"epoch": 0.05,
"eval_validation_agenda_digitale_loss": 1.5326132774353027,
"eval_validation_agenda_digitale_runtime": 178.5645,
"eval_validation_agenda_digitale_samples_per_second": 1.904,
"eval_validation_agenda_digitale_steps_per_second": 1.904,
"step": 750
},
{
"epoch": 0.05,
"eval_validation_leggepertutti_loss": 1.3737919330596924,
"eval_validation_leggepertutti_runtime": 34.1417,
"eval_validation_leggepertutti_samples_per_second": 1.904,
"eval_validation_leggepertutti_steps_per_second": 1.904,
"step": 750
},
{
"epoch": 0.05,
"eval_validation_altalexprivacy_loss": 1.3635202646255493,
"eval_validation_altalexprivacy_runtime": 55.6795,
"eval_validation_altalexprivacy_samples_per_second": 1.904,
"eval_validation_altalexprivacy_steps_per_second": 1.904,
"step": 750
},
{
"epoch": 0.05,
"eval_validation_mc4_it_loss": 1.9601805210113525,
"eval_validation_mc4_it_runtime": 74.5661,
"eval_validation_mc4_it_samples_per_second": 1.904,
"eval_validation_mc4_it_steps_per_second": 1.904,
"step": 750
},
{
"epoch": 0.07,
"learning_rate": 1e-05,
"loss": 1.5963,
"step": 1125
},
{
"epoch": 0.07,
"eval_validation_privacy_sources_loss": 0.9522767066955566,
"eval_validation_privacy_sources_runtime": 201.0521,
"eval_validation_privacy_sources_samples_per_second": 1.905,
"eval_validation_privacy_sources_steps_per_second": 1.905,
"step": 1125
},
{
"epoch": 0.07,
"eval_validation_agenda_digitale_loss": 1.4975751638412476,
"eval_validation_agenda_digitale_runtime": 178.5472,
"eval_validation_agenda_digitale_samples_per_second": 1.904,
"eval_validation_agenda_digitale_steps_per_second": 1.904,
"step": 1125
},
{
"epoch": 0.07,
"eval_validation_leggepertutti_loss": 1.3494906425476074,
"eval_validation_leggepertutti_runtime": 34.1339,
"eval_validation_leggepertutti_samples_per_second": 1.904,
"eval_validation_leggepertutti_steps_per_second": 1.904,
"step": 1125
},
{
"epoch": 0.07,
"eval_validation_altalexprivacy_loss": 1.3334871530532837,
"eval_validation_altalexprivacy_runtime": 55.6542,
"eval_validation_altalexprivacy_samples_per_second": 1.905,
"eval_validation_altalexprivacy_steps_per_second": 1.905,
"step": 1125
},
{
"epoch": 0.07,
"eval_validation_mc4_it_loss": 1.9144717454910278,
"eval_validation_mc4_it_runtime": 74.5606,
"eval_validation_mc4_it_samples_per_second": 1.904,
"eval_validation_mc4_it_steps_per_second": 1.904,
"step": 1125
},
{
"epoch": 0.09,
"learning_rate": 1e-05,
"loss": 1.577,
"step": 1500
},
{
"epoch": 0.09,
"eval_validation_privacy_sources_loss": 0.9246706366539001,
"eval_validation_privacy_sources_runtime": 201.0413,
"eval_validation_privacy_sources_samples_per_second": 1.905,
"eval_validation_privacy_sources_steps_per_second": 1.905,
"step": 1500
},
{
"epoch": 0.09,
"eval_validation_agenda_digitale_loss": 1.4834680557250977,
"eval_validation_agenda_digitale_runtime": 178.5955,
"eval_validation_agenda_digitale_samples_per_second": 1.904,
"eval_validation_agenda_digitale_steps_per_second": 1.904,
"step": 1500
},
{
"epoch": 0.09,
"eval_validation_leggepertutti_loss": 1.3424042463302612,
"eval_validation_leggepertutti_runtime": 34.1348,
"eval_validation_leggepertutti_samples_per_second": 1.904,
"eval_validation_leggepertutti_steps_per_second": 1.904,
"step": 1500
},
{
"epoch": 0.09,
"eval_validation_altalexprivacy_loss": 1.3129479885101318,
"eval_validation_altalexprivacy_runtime": 55.6818,
"eval_validation_altalexprivacy_samples_per_second": 1.904,
"eval_validation_altalexprivacy_steps_per_second": 1.904,
"step": 1500
},
{
"epoch": 0.09,
"eval_validation_mc4_it_loss": 1.897855281829834,
"eval_validation_mc4_it_runtime": 74.5687,
"eval_validation_mc4_it_samples_per_second": 1.904,
"eval_validation_mc4_it_steps_per_second": 1.904,
"step": 1500
},
{
"epoch": 0.11,
"learning_rate": 1e-05,
"loss": 1.551,
"step": 1875
},
{
"epoch": 0.11,
"eval_validation_privacy_sources_loss": 0.9161506295204163,
"eval_validation_privacy_sources_runtime": 201.028,
"eval_validation_privacy_sources_samples_per_second": 1.905,
"eval_validation_privacy_sources_steps_per_second": 1.905,
"step": 1875
},
{
"epoch": 0.11,
"eval_validation_agenda_digitale_loss": 1.4714491367340088,
"eval_validation_agenda_digitale_runtime": 178.5046,
"eval_validation_agenda_digitale_samples_per_second": 1.905,
"eval_validation_agenda_digitale_steps_per_second": 1.905,
"step": 1875
},
{
"epoch": 0.11,
"eval_validation_leggepertutti_loss": 1.349782109260559,
"eval_validation_leggepertutti_runtime": 34.1218,
"eval_validation_leggepertutti_samples_per_second": 1.905,
"eval_validation_leggepertutti_steps_per_second": 1.905,
"step": 1875
},
{
"epoch": 0.11,
"eval_validation_altalexprivacy_loss": 1.3038463592529297,
"eval_validation_altalexprivacy_runtime": 55.6592,
"eval_validation_altalexprivacy_samples_per_second": 1.904,
"eval_validation_altalexprivacy_steps_per_second": 1.904,
"step": 1875
},
{
"epoch": 0.11,
"eval_validation_mc4_it_loss": 1.8886678218841553,
"eval_validation_mc4_it_runtime": 74.5616,
"eval_validation_mc4_it_samples_per_second": 1.904,
"eval_validation_mc4_it_steps_per_second": 1.904,
"step": 1875
},
{
"epoch": 0.14,
"learning_rate": 1e-05,
"loss": 1.5334,
"step": 2250
},
{
"epoch": 0.14,
"eval_validation_privacy_sources_loss": 0.8999343514442444,
"eval_validation_privacy_sources_runtime": 201.0719,
"eval_validation_privacy_sources_samples_per_second": 1.905,
"eval_validation_privacy_sources_steps_per_second": 1.905,
"step": 2250
},
{
"epoch": 0.14,
"eval_validation_agenda_digitale_loss": 1.470467448234558,
"eval_validation_agenda_digitale_runtime": 178.5136,
"eval_validation_agenda_digitale_samples_per_second": 1.905,
"eval_validation_agenda_digitale_steps_per_second": 1.905,
"step": 2250
},
{
"epoch": 0.14,
"eval_validation_leggepertutti_loss": 1.3344582319259644,
"eval_validation_leggepertutti_runtime": 34.1414,
"eval_validation_leggepertutti_samples_per_second": 1.904,
"eval_validation_leggepertutti_steps_per_second": 1.904,
"step": 2250
},
{
"epoch": 0.14,
"eval_validation_altalexprivacy_loss": 1.2899404764175415,
"eval_validation_altalexprivacy_runtime": 55.6737,
"eval_validation_altalexprivacy_samples_per_second": 1.904,
"eval_validation_altalexprivacy_steps_per_second": 1.904,
"step": 2250
},
{
"epoch": 0.14,
"eval_validation_mc4_it_loss": 1.8787847757339478,
"eval_validation_mc4_it_runtime": 74.5812,
"eval_validation_mc4_it_samples_per_second": 1.904,
"eval_validation_mc4_it_steps_per_second": 1.904,
"step": 2250
},
{
"epoch": 0.16,
"learning_rate": 1e-05,
"loss": 1.5289,
"step": 2625
},
{
"epoch": 0.16,
"eval_validation_privacy_sources_loss": 0.8850609064102173,
"eval_validation_privacy_sources_runtime": 201.0584,
"eval_validation_privacy_sources_samples_per_second": 1.905,
"eval_validation_privacy_sources_steps_per_second": 1.905,
"step": 2625
},
{
"epoch": 0.16,
"eval_validation_agenda_digitale_loss": 1.45303213596344,
"eval_validation_agenda_digitale_runtime": 178.5557,
"eval_validation_agenda_digitale_samples_per_second": 1.904,
"eval_validation_agenda_digitale_steps_per_second": 1.904,
"step": 2625
},
{
"epoch": 0.16,
"eval_validation_leggepertutti_loss": 1.3301901817321777,
"eval_validation_leggepertutti_runtime": 34.1396,
"eval_validation_leggepertutti_samples_per_second": 1.904,
"eval_validation_leggepertutti_steps_per_second": 1.904,
"step": 2625
},
{
"epoch": 0.16,
"eval_validation_altalexprivacy_loss": 1.2740957736968994,
"eval_validation_altalexprivacy_runtime": 55.6629,
"eval_validation_altalexprivacy_samples_per_second": 1.904,
"eval_validation_altalexprivacy_steps_per_second": 1.904,
"step": 2625
},
{
"epoch": 0.16,
"eval_validation_mc4_it_loss": 1.8664617538452148,
"eval_validation_mc4_it_runtime": 74.5651,
"eval_validation_mc4_it_samples_per_second": 1.904,
"eval_validation_mc4_it_steps_per_second": 1.904,
"step": 2625
},
{
"epoch": 0.18,
"learning_rate": 1e-05,
"loss": 1.5137,
"step": 3000
},
{
"epoch": 0.18,
"eval_validation_privacy_sources_loss": 0.8736271858215332,
"eval_validation_privacy_sources_runtime": 201.0614,
"eval_validation_privacy_sources_samples_per_second": 1.905,
"eval_validation_privacy_sources_steps_per_second": 1.905,
"step": 3000
},
{
"epoch": 0.18,
"eval_validation_agenda_digitale_loss": 1.4448249340057373,
"eval_validation_agenda_digitale_runtime": 178.5096,
"eval_validation_agenda_digitale_samples_per_second": 1.905,
"eval_validation_agenda_digitale_steps_per_second": 1.905,
"step": 3000
},
{
"epoch": 0.18,
"eval_validation_leggepertutti_loss": 1.3264199495315552,
"eval_validation_leggepertutti_runtime": 34.1392,
"eval_validation_leggepertutti_samples_per_second": 1.904,
"eval_validation_leggepertutti_steps_per_second": 1.904,
"step": 3000
},
{
"epoch": 0.18,
"eval_validation_altalexprivacy_loss": 1.2583509683609009,
"eval_validation_altalexprivacy_runtime": 55.6465,
"eval_validation_altalexprivacy_samples_per_second": 1.905,
"eval_validation_altalexprivacy_steps_per_second": 1.905,
"step": 3000
},
{
"epoch": 0.18,
"eval_validation_mc4_it_loss": 1.8547219038009644,
"eval_validation_mc4_it_runtime": 74.5381,
"eval_validation_mc4_it_samples_per_second": 1.905,
"eval_validation_mc4_it_steps_per_second": 1.905,
"step": 3000
},
{
"epoch": 0.2,
"learning_rate": 1e-05,
"loss": 1.5191,
"step": 3375
},
{
"epoch": 0.2,
"eval_validation_privacy_sources_loss": 0.8653725385665894,
"eval_validation_privacy_sources_runtime": 201.0607,
"eval_validation_privacy_sources_samples_per_second": 1.905,
"eval_validation_privacy_sources_steps_per_second": 1.905,
"step": 3375
},
{
"epoch": 0.2,
"eval_validation_agenda_digitale_loss": 1.437347173690796,
"eval_validation_agenda_digitale_runtime": 178.5114,
"eval_validation_agenda_digitale_samples_per_second": 1.905,
"eval_validation_agenda_digitale_steps_per_second": 1.905,
"step": 3375
},
{
"epoch": 0.2,
"eval_validation_leggepertutti_loss": 1.3095436096191406,
"eval_validation_leggepertutti_runtime": 34.1234,
"eval_validation_leggepertutti_samples_per_second": 1.905,
"eval_validation_leggepertutti_steps_per_second": 1.905,
"step": 3375
},
{
"epoch": 0.2,
"eval_validation_altalexprivacy_loss": 1.24440336227417,
"eval_validation_altalexprivacy_runtime": 55.6399,
"eval_validation_altalexprivacy_samples_per_second": 1.905,
"eval_validation_altalexprivacy_steps_per_second": 1.905,
"step": 3375
},
{
"epoch": 0.2,
"eval_validation_mc4_it_loss": 1.8478875160217285,
"eval_validation_mc4_it_runtime": 74.5544,
"eval_validation_mc4_it_samples_per_second": 1.905,
"eval_validation_mc4_it_steps_per_second": 1.905,
"step": 3375
},
{
"epoch": 0.23,
"learning_rate": 1e-05,
"loss": 1.5014,
"step": 3750
},
{
"epoch": 0.23,
"eval_validation_privacy_sources_loss": 0.866100013256073,
"eval_validation_privacy_sources_runtime": 201.0279,
"eval_validation_privacy_sources_samples_per_second": 1.905,
"eval_validation_privacy_sources_steps_per_second": 1.905,
"step": 3750
},
{
"epoch": 0.23,
"eval_validation_agenda_digitale_loss": 1.4383131265640259,
"eval_validation_agenda_digitale_runtime": 178.5206,
"eval_validation_agenda_digitale_samples_per_second": 1.905,
"eval_validation_agenda_digitale_steps_per_second": 1.905,
"step": 3750
},
{
"epoch": 0.23,
"eval_validation_leggepertutti_loss": 1.3121819496154785,
"eval_validation_leggepertutti_runtime": 34.1311,
"eval_validation_leggepertutti_samples_per_second": 1.904,
"eval_validation_leggepertutti_steps_per_second": 1.904,
"step": 3750
},
{
"epoch": 0.23,
"eval_validation_altalexprivacy_loss": 1.2500646114349365,
"eval_validation_altalexprivacy_runtime": 55.6505,
"eval_validation_altalexprivacy_samples_per_second": 1.905,
"eval_validation_altalexprivacy_steps_per_second": 1.905,
"step": 3750
},
{
"epoch": 0.23,
"eval_validation_mc4_it_loss": 1.8457143306732178,
"eval_validation_mc4_it_runtime": 74.5581,
"eval_validation_mc4_it_samples_per_second": 1.905,
"eval_validation_mc4_it_steps_per_second": 1.905,
"step": 3750
},
{
"epoch": 0.25,
"learning_rate": 1e-05,
"loss": 1.4925,
"step": 4125
},
{
"epoch": 0.25,
"eval_validation_privacy_sources_loss": 0.8534922003746033,
"eval_validation_privacy_sources_runtime": 201.0002,
"eval_validation_privacy_sources_samples_per_second": 1.905,
"eval_validation_privacy_sources_steps_per_second": 1.905,
"step": 4125
},
{
"epoch": 0.25,
"eval_validation_agenda_digitale_loss": 1.426421046257019,
"eval_validation_agenda_digitale_runtime": 178.5283,
"eval_validation_agenda_digitale_samples_per_second": 1.904,
"eval_validation_agenda_digitale_steps_per_second": 1.904,
"step": 4125
},
{
"epoch": 0.25,
"eval_validation_leggepertutti_loss": 1.3124033212661743,
"eval_validation_leggepertutti_runtime": 34.1225,
"eval_validation_leggepertutti_samples_per_second": 1.905,
"eval_validation_leggepertutti_steps_per_second": 1.905,
"step": 4125
},
{
"epoch": 0.25,
"eval_validation_altalexprivacy_loss": 1.2354837656021118,
"eval_validation_altalexprivacy_runtime": 55.6534,
"eval_validation_altalexprivacy_samples_per_second": 1.905,
"eval_validation_altalexprivacy_steps_per_second": 1.905,
"step": 4125
},
{
"epoch": 0.25,
"eval_validation_mc4_it_loss": 1.839884638786316,
"eval_validation_mc4_it_runtime": 74.5669,
"eval_validation_mc4_it_samples_per_second": 1.904,
"eval_validation_mc4_it_steps_per_second": 1.904,
"step": 4125
},
{
"epoch": 0.27,
"learning_rate": 1e-05,
"loss": 1.4953,
"step": 4500
},
{
"epoch": 0.27,
"eval_validation_privacy_sources_loss": 0.849433422088623,
"eval_validation_privacy_sources_runtime": 201.0776,
"eval_validation_privacy_sources_samples_per_second": 1.905,
"eval_validation_privacy_sources_steps_per_second": 1.905,
"step": 4500
},
{
"epoch": 0.27,
"eval_validation_agenda_digitale_loss": 1.4208179712295532,
"eval_validation_agenda_digitale_runtime": 178.5663,
"eval_validation_agenda_digitale_samples_per_second": 1.904,
"eval_validation_agenda_digitale_steps_per_second": 1.904,
"step": 4500
},
{
"epoch": 0.27,
"eval_validation_leggepertutti_loss": 1.3129136562347412,
"eval_validation_leggepertutti_runtime": 34.1427,
"eval_validation_leggepertutti_samples_per_second": 1.904,
"eval_validation_leggepertutti_steps_per_second": 1.904,
"step": 4500
},
{
"epoch": 0.27,
"eval_validation_altalexprivacy_loss": 1.2299654483795166,
"eval_validation_altalexprivacy_runtime": 55.664,
"eval_validation_altalexprivacy_samples_per_second": 1.904,
"eval_validation_altalexprivacy_steps_per_second": 1.904,
"step": 4500
},
{
"epoch": 0.27,
"eval_validation_mc4_it_loss": 1.8318015336990356,
"eval_validation_mc4_it_runtime": 74.5767,
"eval_validation_mc4_it_samples_per_second": 1.904,
"eval_validation_mc4_it_steps_per_second": 1.904,
"step": 4500
},
{
"epoch": 0.29,
"learning_rate": 1e-05,
"loss": 1.4874,
"step": 4875
},
{
"epoch": 0.29,
"eval_validation_privacy_sources_loss": 0.841759979724884,
"eval_validation_privacy_sources_runtime": 201.1289,
"eval_validation_privacy_sources_samples_per_second": 1.904,
"eval_validation_privacy_sources_steps_per_second": 1.904,
"step": 4875
},
{
"epoch": 0.29,
"eval_validation_agenda_digitale_loss": 1.416318655014038,
"eval_validation_agenda_digitale_runtime": 178.5543,
"eval_validation_agenda_digitale_samples_per_second": 1.904,
"eval_validation_agenda_digitale_steps_per_second": 1.904,
"step": 4875
},
{
"epoch": 0.29,
"eval_validation_leggepertutti_loss": 1.3116254806518555,
"eval_validation_leggepertutti_runtime": 34.142,
"eval_validation_leggepertutti_samples_per_second": 1.904,
"eval_validation_leggepertutti_steps_per_second": 1.904,
"step": 4875
},
{
"epoch": 0.29,
"eval_validation_altalexprivacy_loss": 1.2154171466827393,
"eval_validation_altalexprivacy_runtime": 55.6852,
"eval_validation_altalexprivacy_samples_per_second": 1.904,
"eval_validation_altalexprivacy_steps_per_second": 1.904,
"step": 4875
},
{
"epoch": 0.29,
"eval_validation_mc4_it_loss": 1.8277841806411743,
"eval_validation_mc4_it_runtime": 74.593,
"eval_validation_mc4_it_samples_per_second": 1.904,
"eval_validation_mc4_it_steps_per_second": 1.904,
"step": 4875
},
{
"epoch": 0.32,
"learning_rate": 1e-05,
"loss": 1.4732,
"step": 5250
},
{
"epoch": 0.32,
"eval_validation_privacy_sources_loss": 0.8383973240852356,
"eval_validation_privacy_sources_runtime": 201.1658,
"eval_validation_privacy_sources_samples_per_second": 1.904,
"eval_validation_privacy_sources_steps_per_second": 1.904,
"step": 5250
},
{
"epoch": 0.32,
"eval_validation_agenda_digitale_loss": 1.4164748191833496,
"eval_validation_agenda_digitale_runtime": 178.5933,
"eval_validation_agenda_digitale_samples_per_second": 1.904,
"eval_validation_agenda_digitale_steps_per_second": 1.904,
"step": 5250
},
{
"epoch": 0.32,
"eval_validation_leggepertutti_loss": 1.306440830230713,
"eval_validation_leggepertutti_runtime": 34.1439,
"eval_validation_leggepertutti_samples_per_second": 1.904,
"eval_validation_leggepertutti_steps_per_second": 1.904,
"step": 5250
},
{
"epoch": 0.32,
"eval_validation_altalexprivacy_loss": 1.218838095664978,
"eval_validation_altalexprivacy_runtime": 55.6909,
"eval_validation_altalexprivacy_samples_per_second": 1.903,
"eval_validation_altalexprivacy_steps_per_second": 1.903,
"step": 5250
},
{
"epoch": 0.32,
"eval_validation_mc4_it_loss": 1.8245536088943481,
"eval_validation_mc4_it_runtime": 74.6076,
"eval_validation_mc4_it_samples_per_second": 1.903,
"eval_validation_mc4_it_steps_per_second": 1.903,
"step": 5250
},
{
"epoch": 0.34,
"learning_rate": 1e-05,
"loss": 1.4802,
"step": 5625
},
{
"epoch": 0.34,
"eval_validation_privacy_sources_loss": 0.8325279951095581,
"eval_validation_privacy_sources_runtime": 201.1527,
"eval_validation_privacy_sources_samples_per_second": 1.904,
"eval_validation_privacy_sources_steps_per_second": 1.904,
"step": 5625
},
{
"epoch": 0.34,
"eval_validation_agenda_digitale_loss": 1.4099715948104858,
"eval_validation_agenda_digitale_runtime": 178.591,
"eval_validation_agenda_digitale_samples_per_second": 1.904,
"eval_validation_agenda_digitale_steps_per_second": 1.904,
"step": 5625
},
{
"epoch": 0.34,
"eval_validation_leggepertutti_loss": 1.30331552028656,
"eval_validation_leggepertutti_runtime": 34.1424,
"eval_validation_leggepertutti_samples_per_second": 1.904,
"eval_validation_leggepertutti_steps_per_second": 1.904,
"step": 5625
},
{
"epoch": 0.34,
"eval_validation_altalexprivacy_loss": 1.2109583616256714,
"eval_validation_altalexprivacy_runtime": 55.6849,
"eval_validation_altalexprivacy_samples_per_second": 1.904,
"eval_validation_altalexprivacy_steps_per_second": 1.904,
"step": 5625
},
{
"epoch": 0.34,
"eval_validation_mc4_it_loss": 1.8209081888198853,
"eval_validation_mc4_it_runtime": 74.6027,
"eval_validation_mc4_it_samples_per_second": 1.903,
"eval_validation_mc4_it_steps_per_second": 1.903,
"step": 5625
},
{
"epoch": 0.36,
"learning_rate": 1e-05,
"loss": 1.4593,
"step": 6000
},
{
"epoch": 0.36,
"eval_validation_privacy_sources_loss": 0.8251636028289795,
"eval_validation_privacy_sources_runtime": 201.1198,
"eval_validation_privacy_sources_samples_per_second": 1.904,
"eval_validation_privacy_sources_steps_per_second": 1.904,
"step": 6000
},
{
"epoch": 0.36,
"eval_validation_agenda_digitale_loss": 1.4080655574798584,
"eval_validation_agenda_digitale_runtime": 178.6071,
"eval_validation_agenda_digitale_samples_per_second": 1.904,
"eval_validation_agenda_digitale_steps_per_second": 1.904,
"step": 6000
},
{
"epoch": 0.36,
"eval_validation_leggepertutti_loss": 1.3091273307800293,
"eval_validation_leggepertutti_runtime": 34.1414,
"eval_validation_leggepertutti_samples_per_second": 1.904,
"eval_validation_leggepertutti_steps_per_second": 1.904,
"step": 6000
},
{
"epoch": 0.36,
"eval_validation_altalexprivacy_loss": 1.2088643312454224,
"eval_validation_altalexprivacy_runtime": 55.6807,
"eval_validation_altalexprivacy_samples_per_second": 1.904,
"eval_validation_altalexprivacy_steps_per_second": 1.904,
"step": 6000
},
{
"epoch": 0.36,
"eval_validation_mc4_it_loss": 1.8172013759613037,
"eval_validation_mc4_it_runtime": 74.5793,
"eval_validation_mc4_it_samples_per_second": 1.904,
"eval_validation_mc4_it_steps_per_second": 1.904,
"step": 6000
},
{
"epoch": 0.39,
"learning_rate": 1e-05,
"loss": 1.4548,
"step": 6375
},
{
"epoch": 0.39,
"eval_validation_privacy_sources_loss": 0.8178727030754089,
"eval_validation_privacy_sources_runtime": 201.1666,
"eval_validation_privacy_sources_samples_per_second": 1.904,
"eval_validation_privacy_sources_steps_per_second": 1.904,
"step": 6375
},
{
"epoch": 0.39,
"eval_validation_agenda_digitale_loss": 1.4031239748001099,
"eval_validation_agenda_digitale_runtime": 178.5499,
"eval_validation_agenda_digitale_samples_per_second": 1.904,
"eval_validation_agenda_digitale_steps_per_second": 1.904,
"step": 6375
},
{
"epoch": 0.39,
"eval_validation_leggepertutti_loss": 1.3012996912002563,
"eval_validation_leggepertutti_runtime": 34.1355,
"eval_validation_leggepertutti_samples_per_second": 1.904,
"eval_validation_leggepertutti_steps_per_second": 1.904,
"step": 6375
},
{
"epoch": 0.39,
"eval_validation_altalexprivacy_loss": 1.2039263248443604,
"eval_validation_altalexprivacy_runtime": 55.6853,
"eval_validation_altalexprivacy_samples_per_second": 1.904,
"eval_validation_altalexprivacy_steps_per_second": 1.904,
"step": 6375
},
{
"epoch": 0.39,
"eval_validation_mc4_it_loss": 1.8110533952713013,
"eval_validation_mc4_it_runtime": 74.5796,
"eval_validation_mc4_it_samples_per_second": 1.904,
"eval_validation_mc4_it_steps_per_second": 1.904,
"step": 6375
}
],
"logging_steps": 375,
"max_steps": 33094,
"num_input_tokens_seen": 0,
"num_train_epochs": 2,
"save_steps": 375,
"total_flos": 1.3368496062726144e+19,
"train_batch_size": 1,
"trial_name": null,
"trial_params": null
}