|
{ |
|
"best_metric": null, |
|
"best_model_checkpoint": null, |
|
"epoch": 0.902445741262868, |
|
"eval_steps": 187, |
|
"global_step": 4301, |
|
"is_hyper_param_search": false, |
|
"is_local_process_zero": true, |
|
"is_world_process_zero": true, |
|
"log_history": [ |
|
{ |
|
"epoch": 0.0, |
|
"grad_norm": 10.4375, |
|
"learning_rate": 2.1276595744680852e-07, |
|
"loss": 1.5104, |
|
"step": 1 |
|
}, |
|
{ |
|
"epoch": 0.04, |
|
"grad_norm": 2.90625, |
|
"learning_rate": 1e-05, |
|
"loss": 1.1784, |
|
"step": 187 |
|
}, |
|
{ |
|
"epoch": 0.04, |
|
"eval_validation_loss": 0.956233024597168, |
|
"eval_validation_runtime": 190.0844, |
|
"eval_validation_samples_per_second": 0.726, |
|
"eval_validation_steps_per_second": 0.726, |
|
"step": 187 |
|
}, |
|
{ |
|
"epoch": 0.04, |
|
"eval_validation_privacy_sources_loss": 0.956233024597168, |
|
"eval_validation_privacy_sources_runtime": 190.1247, |
|
"eval_validation_privacy_sources_samples_per_second": 0.726, |
|
"eval_validation_privacy_sources_steps_per_second": 0.726, |
|
"step": 187 |
|
}, |
|
{ |
|
"epoch": 0.04, |
|
"eval_validation_agenda_digitale_loss": 1.5249053239822388, |
|
"eval_validation_agenda_digitale_runtime": 170.8208, |
|
"eval_validation_agenda_digitale_samples_per_second": 0.726, |
|
"eval_validation_agenda_digitale_steps_per_second": 0.726, |
|
"step": 187 |
|
}, |
|
{ |
|
"epoch": 0.04, |
|
"eval_validation_legal_articles_loss": 1.4138059616088867, |
|
"eval_validation_legal_articles_runtime": 274.0975, |
|
"eval_validation_legal_articles_samples_per_second": 0.726, |
|
"eval_validation_legal_articles_steps_per_second": 0.726, |
|
"step": 187 |
|
}, |
|
{ |
|
"epoch": 0.04, |
|
"eval_validation_leggepertutti_loss": 1.2507734298706055, |
|
"eval_validation_leggepertutti_runtime": 26.1774, |
|
"eval_validation_leggepertutti_samples_per_second": 0.726, |
|
"eval_validation_leggepertutti_steps_per_second": 0.726, |
|
"step": 187 |
|
}, |
|
{ |
|
"epoch": 0.08, |
|
"grad_norm": 3.5625, |
|
"learning_rate": 1e-05, |
|
"loss": 1.0602, |
|
"step": 374 |
|
}, |
|
{ |
|
"epoch": 0.08, |
|
"eval_validation_loss": 0.9068162441253662, |
|
"eval_validation_runtime": 190.0623, |
|
"eval_validation_samples_per_second": 0.726, |
|
"eval_validation_steps_per_second": 0.726, |
|
"step": 374 |
|
}, |
|
{ |
|
"epoch": 0.08, |
|
"eval_validation_privacy_sources_loss": 0.9068162441253662, |
|
"eval_validation_privacy_sources_runtime": 190.1749, |
|
"eval_validation_privacy_sources_samples_per_second": 0.726, |
|
"eval_validation_privacy_sources_steps_per_second": 0.726, |
|
"step": 374 |
|
}, |
|
{ |
|
"epoch": 0.08, |
|
"eval_validation_agenda_digitale_loss": 1.5025277137756348, |
|
"eval_validation_agenda_digitale_runtime": 170.8116, |
|
"eval_validation_agenda_digitale_samples_per_second": 0.726, |
|
"eval_validation_agenda_digitale_steps_per_second": 0.726, |
|
"step": 374 |
|
}, |
|
{ |
|
"epoch": 0.08, |
|
"eval_validation_legal_articles_loss": 1.3901658058166504, |
|
"eval_validation_legal_articles_runtime": 274.2187, |
|
"eval_validation_legal_articles_samples_per_second": 0.726, |
|
"eval_validation_legal_articles_steps_per_second": 0.726, |
|
"step": 374 |
|
}, |
|
{ |
|
"epoch": 0.08, |
|
"eval_validation_leggepertutti_loss": 1.2389607429504395, |
|
"eval_validation_leggepertutti_runtime": 26.1823, |
|
"eval_validation_leggepertutti_samples_per_second": 0.726, |
|
"eval_validation_leggepertutti_steps_per_second": 0.726, |
|
"step": 374 |
|
}, |
|
{ |
|
"epoch": 0.12, |
|
"grad_norm": 2.375, |
|
"learning_rate": 1e-05, |
|
"loss": 0.9973, |
|
"step": 561 |
|
}, |
|
{ |
|
"epoch": 0.12, |
|
"eval_validation_loss": 0.8726317286491394, |
|
"eval_validation_runtime": 190.0376, |
|
"eval_validation_samples_per_second": 0.726, |
|
"eval_validation_steps_per_second": 0.726, |
|
"step": 561 |
|
}, |
|
{ |
|
"epoch": 0.12, |
|
"eval_validation_privacy_sources_loss": 0.8726317286491394, |
|
"eval_validation_privacy_sources_runtime": 189.9977, |
|
"eval_validation_privacy_sources_samples_per_second": 0.726, |
|
"eval_validation_privacy_sources_steps_per_second": 0.726, |
|
"step": 561 |
|
}, |
|
{ |
|
"epoch": 0.12, |
|
"eval_validation_agenda_digitale_loss": 1.4876974821090698, |
|
"eval_validation_agenda_digitale_runtime": 170.7148, |
|
"eval_validation_agenda_digitale_samples_per_second": 0.726, |
|
"eval_validation_agenda_digitale_steps_per_second": 0.726, |
|
"step": 561 |
|
}, |
|
{ |
|
"epoch": 0.12, |
|
"eval_validation_legal_articles_loss": 1.3721040487289429, |
|
"eval_validation_legal_articles_runtime": 273.9951, |
|
"eval_validation_legal_articles_samples_per_second": 0.726, |
|
"eval_validation_legal_articles_steps_per_second": 0.726, |
|
"step": 561 |
|
}, |
|
{ |
|
"epoch": 0.12, |
|
"eval_validation_leggepertutti_loss": 1.213807463645935, |
|
"eval_validation_leggepertutti_runtime": 26.1576, |
|
"eval_validation_leggepertutti_samples_per_second": 0.726, |
|
"eval_validation_leggepertutti_steps_per_second": 0.726, |
|
"step": 561 |
|
}, |
|
{ |
|
"epoch": 0.16, |
|
"grad_norm": 2.734375, |
|
"learning_rate": 1e-05, |
|
"loss": 0.9936, |
|
"step": 748 |
|
}, |
|
{ |
|
"epoch": 0.16, |
|
"eval_validation_loss": 0.8519095778465271, |
|
"eval_validation_runtime": 190.0914, |
|
"eval_validation_samples_per_second": 0.726, |
|
"eval_validation_steps_per_second": 0.726, |
|
"step": 748 |
|
}, |
|
{ |
|
"epoch": 0.16, |
|
"eval_validation_privacy_sources_loss": 0.8519095778465271, |
|
"eval_validation_privacy_sources_runtime": 190.1871, |
|
"eval_validation_privacy_sources_samples_per_second": 0.726, |
|
"eval_validation_privacy_sources_steps_per_second": 0.726, |
|
"step": 748 |
|
}, |
|
{ |
|
"epoch": 0.16, |
|
"eval_validation_agenda_digitale_loss": 1.4739274978637695, |
|
"eval_validation_agenda_digitale_runtime": 170.8961, |
|
"eval_validation_agenda_digitale_samples_per_second": 0.726, |
|
"eval_validation_agenda_digitale_steps_per_second": 0.726, |
|
"step": 748 |
|
}, |
|
{ |
|
"epoch": 0.16, |
|
"eval_validation_legal_articles_loss": 1.3598031997680664, |
|
"eval_validation_legal_articles_runtime": 274.1856, |
|
"eval_validation_legal_articles_samples_per_second": 0.726, |
|
"eval_validation_legal_articles_steps_per_second": 0.726, |
|
"step": 748 |
|
}, |
|
{ |
|
"epoch": 0.16, |
|
"eval_validation_leggepertutti_loss": 1.2074390649795532, |
|
"eval_validation_leggepertutti_runtime": 26.177, |
|
"eval_validation_leggepertutti_samples_per_second": 0.726, |
|
"eval_validation_leggepertutti_steps_per_second": 0.726, |
|
"step": 748 |
|
}, |
|
{ |
|
"epoch": 0.2, |
|
"grad_norm": 2.90625, |
|
"learning_rate": 1e-05, |
|
"loss": 0.9671, |
|
"step": 935 |
|
}, |
|
{ |
|
"epoch": 0.2, |
|
"eval_validation_loss": 0.8343605399131775, |
|
"eval_validation_runtime": 190.0578, |
|
"eval_validation_samples_per_second": 0.726, |
|
"eval_validation_steps_per_second": 0.726, |
|
"step": 935 |
|
}, |
|
{ |
|
"epoch": 0.2, |
|
"eval_validation_privacy_sources_loss": 0.8343605399131775, |
|
"eval_validation_privacy_sources_runtime": 190.0981, |
|
"eval_validation_privacy_sources_samples_per_second": 0.726, |
|
"eval_validation_privacy_sources_steps_per_second": 0.726, |
|
"step": 935 |
|
}, |
|
{ |
|
"epoch": 0.2, |
|
"eval_validation_agenda_digitale_loss": 1.4663511514663696, |
|
"eval_validation_agenda_digitale_runtime": 170.7988, |
|
"eval_validation_agenda_digitale_samples_per_second": 0.726, |
|
"eval_validation_agenda_digitale_steps_per_second": 0.726, |
|
"step": 935 |
|
}, |
|
{ |
|
"epoch": 0.2, |
|
"eval_validation_legal_articles_loss": 1.3504709005355835, |
|
"eval_validation_legal_articles_runtime": 274.1096, |
|
"eval_validation_legal_articles_samples_per_second": 0.726, |
|
"eval_validation_legal_articles_steps_per_second": 0.726, |
|
"step": 935 |
|
}, |
|
{ |
|
"epoch": 0.2, |
|
"eval_validation_leggepertutti_loss": 1.19623863697052, |
|
"eval_validation_leggepertutti_runtime": 26.1751, |
|
"eval_validation_leggepertutti_samples_per_second": 0.726, |
|
"eval_validation_leggepertutti_steps_per_second": 0.726, |
|
"step": 935 |
|
}, |
|
{ |
|
"epoch": 0.24, |
|
"grad_norm": 18.625, |
|
"learning_rate": 1e-05, |
|
"loss": 0.9582, |
|
"step": 1122 |
|
}, |
|
{ |
|
"epoch": 0.24, |
|
"eval_validation_loss": 0.8237400650978088, |
|
"eval_validation_runtime": 190.0819, |
|
"eval_validation_samples_per_second": 0.726, |
|
"eval_validation_steps_per_second": 0.726, |
|
"step": 1122 |
|
}, |
|
{ |
|
"epoch": 0.24, |
|
"eval_validation_privacy_sources_loss": 0.8237400650978088, |
|
"eval_validation_privacy_sources_runtime": 190.0975, |
|
"eval_validation_privacy_sources_samples_per_second": 0.726, |
|
"eval_validation_privacy_sources_steps_per_second": 0.726, |
|
"step": 1122 |
|
}, |
|
{ |
|
"epoch": 0.24, |
|
"eval_validation_agenda_digitale_loss": 1.4592102766036987, |
|
"eval_validation_agenda_digitale_runtime": 170.7961, |
|
"eval_validation_agenda_digitale_samples_per_second": 0.726, |
|
"eval_validation_agenda_digitale_steps_per_second": 0.726, |
|
"step": 1122 |
|
}, |
|
{ |
|
"epoch": 0.24, |
|
"eval_validation_legal_articles_loss": 1.341143250465393, |
|
"eval_validation_legal_articles_runtime": 274.155, |
|
"eval_validation_legal_articles_samples_per_second": 0.726, |
|
"eval_validation_legal_articles_steps_per_second": 0.726, |
|
"step": 1122 |
|
}, |
|
{ |
|
"epoch": 0.24, |
|
"eval_validation_leggepertutti_loss": 1.1890021562576294, |
|
"eval_validation_leggepertutti_runtime": 26.1699, |
|
"eval_validation_leggepertutti_samples_per_second": 0.726, |
|
"eval_validation_leggepertutti_steps_per_second": 0.726, |
|
"step": 1122 |
|
}, |
|
{ |
|
"epoch": 0.27, |
|
"grad_norm": 2.28125, |
|
"learning_rate": 1e-05, |
|
"loss": 0.9456, |
|
"step": 1309 |
|
}, |
|
{ |
|
"epoch": 0.27, |
|
"eval_validation_loss": 0.8131315112113953, |
|
"eval_validation_runtime": 189.7603, |
|
"eval_validation_samples_per_second": 0.727, |
|
"eval_validation_steps_per_second": 0.727, |
|
"step": 1309 |
|
}, |
|
{ |
|
"epoch": 0.27, |
|
"eval_validation_privacy_sources_loss": 0.8131315112113953, |
|
"eval_validation_privacy_sources_runtime": 189.8097, |
|
"eval_validation_privacy_sources_samples_per_second": 0.727, |
|
"eval_validation_privacy_sources_steps_per_second": 0.727, |
|
"step": 1309 |
|
}, |
|
{ |
|
"epoch": 0.27, |
|
"eval_validation_agenda_digitale_loss": 1.4516360759735107, |
|
"eval_validation_agenda_digitale_runtime": 170.4955, |
|
"eval_validation_agenda_digitale_samples_per_second": 0.727, |
|
"eval_validation_agenda_digitale_steps_per_second": 0.727, |
|
"step": 1309 |
|
}, |
|
{ |
|
"epoch": 0.27, |
|
"eval_validation_legal_articles_loss": 1.3339463472366333, |
|
"eval_validation_legal_articles_runtime": 273.5945, |
|
"eval_validation_legal_articles_samples_per_second": 0.727, |
|
"eval_validation_legal_articles_steps_per_second": 0.727, |
|
"step": 1309 |
|
}, |
|
{ |
|
"epoch": 0.27, |
|
"eval_validation_leggepertutti_loss": 1.1844114065170288, |
|
"eval_validation_leggepertutti_runtime": 26.1031, |
|
"eval_validation_leggepertutti_samples_per_second": 0.728, |
|
"eval_validation_leggepertutti_steps_per_second": 0.728, |
|
"step": 1309 |
|
}, |
|
{ |
|
"epoch": 0.31, |
|
"grad_norm": 2.203125, |
|
"learning_rate": 1e-05, |
|
"loss": 0.9118, |
|
"step": 1496 |
|
}, |
|
{ |
|
"epoch": 0.31, |
|
"eval_validation_loss": 0.8056552410125732, |
|
"eval_validation_runtime": 189.1065, |
|
"eval_validation_samples_per_second": 0.73, |
|
"eval_validation_steps_per_second": 0.73, |
|
"step": 1496 |
|
}, |
|
{ |
|
"epoch": 0.31, |
|
"eval_validation_privacy_sources_loss": 0.8056552410125732, |
|
"eval_validation_privacy_sources_runtime": 189.1418, |
|
"eval_validation_privacy_sources_samples_per_second": 0.73, |
|
"eval_validation_privacy_sources_steps_per_second": 0.73, |
|
"step": 1496 |
|
}, |
|
{ |
|
"epoch": 0.31, |
|
"eval_validation_agenda_digitale_loss": 1.4482883214950562, |
|
"eval_validation_agenda_digitale_runtime": 169.9314, |
|
"eval_validation_agenda_digitale_samples_per_second": 0.73, |
|
"eval_validation_agenda_digitale_steps_per_second": 0.73, |
|
"step": 1496 |
|
}, |
|
{ |
|
"epoch": 0.31, |
|
"eval_validation_legal_articles_loss": 1.3298695087432861, |
|
"eval_validation_legal_articles_runtime": 272.7633, |
|
"eval_validation_legal_articles_samples_per_second": 0.73, |
|
"eval_validation_legal_articles_steps_per_second": 0.73, |
|
"step": 1496 |
|
}, |
|
{ |
|
"epoch": 0.31, |
|
"eval_validation_leggepertutti_loss": 1.178873896598816, |
|
"eval_validation_leggepertutti_runtime": 26.0447, |
|
"eval_validation_leggepertutti_samples_per_second": 0.73, |
|
"eval_validation_leggepertutti_steps_per_second": 0.73, |
|
"step": 1496 |
|
}, |
|
{ |
|
"epoch": 0.35, |
|
"grad_norm": 2.28125, |
|
"learning_rate": 1e-05, |
|
"loss": 0.9018, |
|
"step": 1683 |
|
}, |
|
{ |
|
"epoch": 0.35, |
|
"eval_validation_loss": 0.7955409288406372, |
|
"eval_validation_runtime": 190.0308, |
|
"eval_validation_samples_per_second": 0.726, |
|
"eval_validation_steps_per_second": 0.726, |
|
"step": 1683 |
|
}, |
|
{ |
|
"epoch": 0.35, |
|
"eval_validation_privacy_sources_loss": 0.7955409288406372, |
|
"eval_validation_privacy_sources_runtime": 190.1015, |
|
"eval_validation_privacy_sources_samples_per_second": 0.726, |
|
"eval_validation_privacy_sources_steps_per_second": 0.726, |
|
"step": 1683 |
|
}, |
|
{ |
|
"epoch": 0.35, |
|
"eval_validation_agenda_digitale_loss": 1.4420207738876343, |
|
"eval_validation_agenda_digitale_runtime": 170.7677, |
|
"eval_validation_agenda_digitale_samples_per_second": 0.726, |
|
"eval_validation_agenda_digitale_steps_per_second": 0.726, |
|
"step": 1683 |
|
}, |
|
{ |
|
"epoch": 0.35, |
|
"eval_validation_legal_articles_loss": 1.3251116275787354, |
|
"eval_validation_legal_articles_runtime": 274.0824, |
|
"eval_validation_legal_articles_samples_per_second": 0.726, |
|
"eval_validation_legal_articles_steps_per_second": 0.726, |
|
"step": 1683 |
|
}, |
|
{ |
|
"epoch": 0.35, |
|
"eval_validation_leggepertutti_loss": 1.1761751174926758, |
|
"eval_validation_leggepertutti_runtime": 26.1786, |
|
"eval_validation_leggepertutti_samples_per_second": 0.726, |
|
"eval_validation_leggepertutti_steps_per_second": 0.726, |
|
"step": 1683 |
|
}, |
|
{ |
|
"epoch": 0.39, |
|
"grad_norm": 3.375, |
|
"learning_rate": 1e-05, |
|
"loss": 0.8911, |
|
"step": 1870 |
|
}, |
|
{ |
|
"epoch": 0.39, |
|
"eval_validation_loss": 0.7886275053024292, |
|
"eval_validation_runtime": 190.0169, |
|
"eval_validation_samples_per_second": 0.726, |
|
"eval_validation_steps_per_second": 0.726, |
|
"step": 1870 |
|
}, |
|
{ |
|
"epoch": 0.39, |
|
"eval_validation_privacy_sources_loss": 0.7886275053024292, |
|
"eval_validation_privacy_sources_runtime": 189.9755, |
|
"eval_validation_privacy_sources_samples_per_second": 0.726, |
|
"eval_validation_privacy_sources_steps_per_second": 0.726, |
|
"step": 1870 |
|
}, |
|
{ |
|
"epoch": 0.39, |
|
"eval_validation_agenda_digitale_loss": 1.437984585762024, |
|
"eval_validation_agenda_digitale_runtime": 170.7137, |
|
"eval_validation_agenda_digitale_samples_per_second": 0.726, |
|
"eval_validation_agenda_digitale_steps_per_second": 0.726, |
|
"step": 1870 |
|
}, |
|
{ |
|
"epoch": 0.39, |
|
"eval_validation_legal_articles_loss": 1.3213441371917725, |
|
"eval_validation_legal_articles_runtime": 273.9792, |
|
"eval_validation_legal_articles_samples_per_second": 0.726, |
|
"eval_validation_legal_articles_steps_per_second": 0.726, |
|
"step": 1870 |
|
}, |
|
{ |
|
"epoch": 0.39, |
|
"eval_validation_leggepertutti_loss": 1.1724990606307983, |
|
"eval_validation_leggepertutti_runtime": 26.1471, |
|
"eval_validation_leggepertutti_samples_per_second": 0.727, |
|
"eval_validation_leggepertutti_steps_per_second": 0.727, |
|
"step": 1870 |
|
}, |
|
{ |
|
"epoch": 0.43, |
|
"grad_norm": 2.34375, |
|
"learning_rate": 1e-05, |
|
"loss": 0.8824, |
|
"step": 2057 |
|
}, |
|
{ |
|
"epoch": 0.43, |
|
"eval_validation_loss": 0.7812964916229248, |
|
"eval_validation_runtime": 189.9852, |
|
"eval_validation_samples_per_second": 0.726, |
|
"eval_validation_steps_per_second": 0.726, |
|
"step": 2057 |
|
}, |
|
{ |
|
"epoch": 0.43, |
|
"eval_validation_privacy_sources_loss": 0.7812964916229248, |
|
"eval_validation_privacy_sources_runtime": 190.0648, |
|
"eval_validation_privacy_sources_samples_per_second": 0.726, |
|
"eval_validation_privacy_sources_steps_per_second": 0.726, |
|
"step": 2057 |
|
}, |
|
{ |
|
"epoch": 0.43, |
|
"eval_validation_agenda_digitale_loss": 1.4352736473083496, |
|
"eval_validation_agenda_digitale_runtime": 170.7674, |
|
"eval_validation_agenda_digitale_samples_per_second": 0.726, |
|
"eval_validation_agenda_digitale_steps_per_second": 0.726, |
|
"step": 2057 |
|
}, |
|
{ |
|
"epoch": 0.43, |
|
"eval_validation_legal_articles_loss": 1.3146668672561646, |
|
"eval_validation_legal_articles_runtime": 274.0088, |
|
"eval_validation_legal_articles_samples_per_second": 0.726, |
|
"eval_validation_legal_articles_steps_per_second": 0.726, |
|
"step": 2057 |
|
}, |
|
{ |
|
"epoch": 0.43, |
|
"eval_validation_leggepertutti_loss": 1.1654175519943237, |
|
"eval_validation_leggepertutti_runtime": 26.1671, |
|
"eval_validation_leggepertutti_samples_per_second": 0.726, |
|
"eval_validation_leggepertutti_steps_per_second": 0.726, |
|
"step": 2057 |
|
}, |
|
{ |
|
"epoch": 0.47, |
|
"grad_norm": 2.515625, |
|
"learning_rate": 1e-05, |
|
"loss": 0.8871, |
|
"step": 2244 |
|
}, |
|
{ |
|
"epoch": 0.47, |
|
"eval_validation_loss": 0.7756069898605347, |
|
"eval_validation_runtime": 190.0643, |
|
"eval_validation_samples_per_second": 0.726, |
|
"eval_validation_steps_per_second": 0.726, |
|
"step": 2244 |
|
}, |
|
{ |
|
"epoch": 0.47, |
|
"eval_validation_privacy_sources_loss": 0.7756069898605347, |
|
"eval_validation_privacy_sources_runtime": 190.0095, |
|
"eval_validation_privacy_sources_samples_per_second": 0.726, |
|
"eval_validation_privacy_sources_steps_per_second": 0.726, |
|
"step": 2244 |
|
}, |
|
{ |
|
"epoch": 0.47, |
|
"eval_validation_agenda_digitale_loss": 1.431518316268921, |
|
"eval_validation_agenda_digitale_runtime": 170.7517, |
|
"eval_validation_agenda_digitale_samples_per_second": 0.726, |
|
"eval_validation_agenda_digitale_steps_per_second": 0.726, |
|
"step": 2244 |
|
}, |
|
{ |
|
"epoch": 0.47, |
|
"eval_validation_legal_articles_loss": 1.3110154867172241, |
|
"eval_validation_legal_articles_runtime": 274.0835, |
|
"eval_validation_legal_articles_samples_per_second": 0.726, |
|
"eval_validation_legal_articles_steps_per_second": 0.726, |
|
"step": 2244 |
|
}, |
|
{ |
|
"epoch": 0.47, |
|
"eval_validation_leggepertutti_loss": 1.161647915840149, |
|
"eval_validation_leggepertutti_runtime": 26.1417, |
|
"eval_validation_leggepertutti_samples_per_second": 0.727, |
|
"eval_validation_leggepertutti_steps_per_second": 0.727, |
|
"step": 2244 |
|
}, |
|
{ |
|
"epoch": 0.51, |
|
"grad_norm": 2.640625, |
|
"learning_rate": 1e-05, |
|
"loss": 0.8723, |
|
"step": 2431 |
|
}, |
|
{ |
|
"epoch": 0.51, |
|
"eval_validation_loss": 0.7698301076889038, |
|
"eval_validation_runtime": 189.9735, |
|
"eval_validation_samples_per_second": 0.726, |
|
"eval_validation_steps_per_second": 0.726, |
|
"step": 2431 |
|
}, |
|
{ |
|
"epoch": 0.51, |
|
"eval_validation_privacy_sources_loss": 0.7698301076889038, |
|
"eval_validation_privacy_sources_runtime": 190.0719, |
|
"eval_validation_privacy_sources_samples_per_second": 0.726, |
|
"eval_validation_privacy_sources_steps_per_second": 0.726, |
|
"step": 2431 |
|
}, |
|
{ |
|
"epoch": 0.51, |
|
"eval_validation_agenda_digitale_loss": 1.4284371137619019, |
|
"eval_validation_agenda_digitale_runtime": 170.7398, |
|
"eval_validation_agenda_digitale_samples_per_second": 0.726, |
|
"eval_validation_agenda_digitale_steps_per_second": 0.726, |
|
"step": 2431 |
|
}, |
|
{ |
|
"epoch": 0.51, |
|
"eval_validation_legal_articles_loss": 1.3073266744613647, |
|
"eval_validation_legal_articles_runtime": 273.9802, |
|
"eval_validation_legal_articles_samples_per_second": 0.726, |
|
"eval_validation_legal_articles_steps_per_second": 0.726, |
|
"step": 2431 |
|
}, |
|
{ |
|
"epoch": 0.51, |
|
"eval_validation_leggepertutti_loss": 1.1593568325042725, |
|
"eval_validation_leggepertutti_runtime": 26.1645, |
|
"eval_validation_leggepertutti_samples_per_second": 0.726, |
|
"eval_validation_leggepertutti_steps_per_second": 0.726, |
|
"step": 2431 |
|
}, |
|
{ |
|
"epoch": 0.55, |
|
"grad_norm": 2.203125, |
|
"learning_rate": 1e-05, |
|
"loss": 0.8663, |
|
"step": 2618 |
|
}, |
|
{ |
|
"epoch": 0.55, |
|
"eval_validation_loss": 0.7643269896507263, |
|
"eval_validation_runtime": 189.9762, |
|
"eval_validation_samples_per_second": 0.726, |
|
"eval_validation_steps_per_second": 0.726, |
|
"step": 2618 |
|
}, |
|
{ |
|
"epoch": 0.55, |
|
"eval_validation_privacy_sources_loss": 0.7643269896507263, |
|
"eval_validation_privacy_sources_runtime": 189.979, |
|
"eval_validation_privacy_sources_samples_per_second": 0.726, |
|
"eval_validation_privacy_sources_steps_per_second": 0.726, |
|
"step": 2618 |
|
}, |
|
{ |
|
"epoch": 0.55, |
|
"eval_validation_agenda_digitale_loss": 1.4276658296585083, |
|
"eval_validation_agenda_digitale_runtime": 170.7641, |
|
"eval_validation_agenda_digitale_samples_per_second": 0.726, |
|
"eval_validation_agenda_digitale_steps_per_second": 0.726, |
|
"step": 2618 |
|
}, |
|
{ |
|
"epoch": 0.55, |
|
"eval_validation_legal_articles_loss": 1.306546926498413, |
|
"eval_validation_legal_articles_runtime": 274.0313, |
|
"eval_validation_legal_articles_samples_per_second": 0.726, |
|
"eval_validation_legal_articles_steps_per_second": 0.726, |
|
"step": 2618 |
|
}, |
|
{ |
|
"epoch": 0.55, |
|
"eval_validation_leggepertutti_loss": 1.1585664749145508, |
|
"eval_validation_leggepertutti_runtime": 26.1644, |
|
"eval_validation_leggepertutti_samples_per_second": 0.726, |
|
"eval_validation_leggepertutti_steps_per_second": 0.726, |
|
"step": 2618 |
|
}, |
|
{ |
|
"epoch": 0.59, |
|
"grad_norm": 2.75, |
|
"learning_rate": 1e-05, |
|
"loss": 0.8765, |
|
"step": 2805 |
|
}, |
|
{ |
|
"epoch": 0.59, |
|
"eval_validation_loss": 0.7591044306755066, |
|
"eval_validation_runtime": 189.882, |
|
"eval_validation_samples_per_second": 0.727, |
|
"eval_validation_steps_per_second": 0.727, |
|
"step": 2805 |
|
}, |
|
{ |
|
"epoch": 0.59, |
|
"eval_validation_privacy_sources_loss": 0.7591044306755066, |
|
"eval_validation_privacy_sources_runtime": 189.9501, |
|
"eval_validation_privacy_sources_samples_per_second": 0.727, |
|
"eval_validation_privacy_sources_steps_per_second": 0.727, |
|
"step": 2805 |
|
}, |
|
{ |
|
"epoch": 0.59, |
|
"eval_validation_agenda_digitale_loss": 1.4212045669555664, |
|
"eval_validation_agenda_digitale_runtime": 170.6266, |
|
"eval_validation_agenda_digitale_samples_per_second": 0.727, |
|
"eval_validation_agenda_digitale_steps_per_second": 0.727, |
|
"step": 2805 |
|
}, |
|
{ |
|
"epoch": 0.59, |
|
"eval_validation_legal_articles_loss": 1.3000119924545288, |
|
"eval_validation_legal_articles_runtime": 273.9772, |
|
"eval_validation_legal_articles_samples_per_second": 0.726, |
|
"eval_validation_legal_articles_steps_per_second": 0.726, |
|
"step": 2805 |
|
}, |
|
{ |
|
"epoch": 0.59, |
|
"eval_validation_leggepertutti_loss": 1.1529566049575806, |
|
"eval_validation_leggepertutti_runtime": 26.1754, |
|
"eval_validation_leggepertutti_samples_per_second": 0.726, |
|
"eval_validation_leggepertutti_steps_per_second": 0.726, |
|
"step": 2805 |
|
}, |
|
{ |
|
"epoch": 0.63, |
|
"grad_norm": 2.25, |
|
"learning_rate": 1e-05, |
|
"loss": 0.8603, |
|
"step": 2992 |
|
}, |
|
{ |
|
"epoch": 0.63, |
|
"eval_validation_loss": 0.7559717893600464, |
|
"eval_validation_runtime": 189.9515, |
|
"eval_validation_samples_per_second": 0.727, |
|
"eval_validation_steps_per_second": 0.727, |
|
"step": 2992 |
|
}, |
|
{ |
|
"epoch": 0.63, |
|
"eval_validation_privacy_sources_loss": 0.7559717893600464, |
|
"eval_validation_privacy_sources_runtime": 189.9552, |
|
"eval_validation_privacy_sources_samples_per_second": 0.726, |
|
"eval_validation_privacy_sources_steps_per_second": 0.726, |
|
"step": 2992 |
|
}, |
|
{ |
|
"epoch": 0.63, |
|
"eval_validation_agenda_digitale_loss": 1.4198051691055298, |
|
"eval_validation_agenda_digitale_runtime": 170.7077, |
|
"eval_validation_agenda_digitale_samples_per_second": 0.726, |
|
"eval_validation_agenda_digitale_steps_per_second": 0.726, |
|
"step": 2992 |
|
}, |
|
{ |
|
"epoch": 0.63, |
|
"eval_validation_legal_articles_loss": 1.2991034984588623, |
|
"eval_validation_legal_articles_runtime": 273.9099, |
|
"eval_validation_legal_articles_samples_per_second": 0.727, |
|
"eval_validation_legal_articles_steps_per_second": 0.727, |
|
"step": 2992 |
|
}, |
|
{ |
|
"epoch": 0.63, |
|
"eval_validation_leggepertutti_loss": 1.1550276279449463, |
|
"eval_validation_leggepertutti_runtime": 26.1726, |
|
"eval_validation_leggepertutti_samples_per_second": 0.726, |
|
"eval_validation_leggepertutti_steps_per_second": 0.726, |
|
"step": 2992 |
|
}, |
|
{ |
|
"epoch": 0.67, |
|
"grad_norm": 2.28125, |
|
"learning_rate": 1e-05, |
|
"loss": 0.8584, |
|
"step": 3179 |
|
}, |
|
{ |
|
"epoch": 0.67, |
|
"eval_validation_loss": 0.7532868385314941, |
|
"eval_validation_runtime": 190.0516, |
|
"eval_validation_samples_per_second": 0.726, |
|
"eval_validation_steps_per_second": 0.726, |
|
"step": 3179 |
|
}, |
|
{ |
|
"epoch": 0.67, |
|
"eval_validation_privacy_sources_loss": 0.7532868385314941, |
|
"eval_validation_privacy_sources_runtime": 190.0874, |
|
"eval_validation_privacy_sources_samples_per_second": 0.726, |
|
"eval_validation_privacy_sources_steps_per_second": 0.726, |
|
"step": 3179 |
|
}, |
|
{ |
|
"epoch": 0.67, |
|
"eval_validation_agenda_digitale_loss": 1.4165968894958496, |
|
"eval_validation_agenda_digitale_runtime": 170.7869, |
|
"eval_validation_agenda_digitale_samples_per_second": 0.726, |
|
"eval_validation_agenda_digitale_steps_per_second": 0.726, |
|
"step": 3179 |
|
}, |
|
{ |
|
"epoch": 0.67, |
|
"eval_validation_legal_articles_loss": 1.2945970296859741, |
|
"eval_validation_legal_articles_runtime": 273.9816, |
|
"eval_validation_legal_articles_samples_per_second": 0.726, |
|
"eval_validation_legal_articles_steps_per_second": 0.726, |
|
"step": 3179 |
|
}, |
|
{ |
|
"epoch": 0.67, |
|
"eval_validation_leggepertutti_loss": 1.1494359970092773, |
|
"eval_validation_leggepertutti_runtime": 26.1538, |
|
"eval_validation_leggepertutti_samples_per_second": 0.726, |
|
"eval_validation_leggepertutti_steps_per_second": 0.726, |
|
"step": 3179 |
|
}, |
|
{ |
|
"epoch": 0.71, |
|
"grad_norm": 10.875, |
|
"learning_rate": 1e-05, |
|
"loss": 0.8506, |
|
"step": 3366 |
|
}, |
|
{ |
|
"epoch": 0.71, |
|
"eval_validation_loss": 0.750731885433197, |
|
"eval_validation_runtime": 190.0222, |
|
"eval_validation_samples_per_second": 0.726, |
|
"eval_validation_steps_per_second": 0.726, |
|
"step": 3366 |
|
}, |
|
{ |
|
"epoch": 0.71, |
|
"eval_validation_privacy_sources_loss": 0.750731885433197, |
|
"eval_validation_privacy_sources_runtime": 190.0188, |
|
"eval_validation_privacy_sources_samples_per_second": 0.726, |
|
"eval_validation_privacy_sources_steps_per_second": 0.726, |
|
"step": 3366 |
|
}, |
|
{ |
|
"epoch": 0.71, |
|
"eval_validation_agenda_digitale_loss": 1.4151290655136108, |
|
"eval_validation_agenda_digitale_runtime": 170.7202, |
|
"eval_validation_agenda_digitale_samples_per_second": 0.726, |
|
"eval_validation_agenda_digitale_steps_per_second": 0.726, |
|
"step": 3366 |
|
}, |
|
{ |
|
"epoch": 0.71, |
|
"eval_validation_legal_articles_loss": 1.2938265800476074, |
|
"eval_validation_legal_articles_runtime": 274.068, |
|
"eval_validation_legal_articles_samples_per_second": 0.726, |
|
"eval_validation_legal_articles_steps_per_second": 0.726, |
|
"step": 3366 |
|
}, |
|
{ |
|
"epoch": 0.71, |
|
"eval_validation_leggepertutti_loss": 1.151176929473877, |
|
"eval_validation_leggepertutti_runtime": 26.1605, |
|
"eval_validation_leggepertutti_samples_per_second": 0.726, |
|
"eval_validation_leggepertutti_steps_per_second": 0.726, |
|
"step": 3366 |
|
}, |
|
{ |
|
"epoch": 0.75, |
|
"grad_norm": 3.0, |
|
"learning_rate": 1e-05, |
|
"loss": 0.8608, |
|
"step": 3553 |
|
}, |
|
{ |
|
"epoch": 0.75, |
|
"eval_validation_loss": 0.7474144101142883, |
|
"eval_validation_runtime": 189.9367, |
|
"eval_validation_samples_per_second": 0.727, |
|
"eval_validation_steps_per_second": 0.727, |
|
"step": 3553 |
|
}, |
|
{ |
|
"epoch": 0.75, |
|
"eval_validation_privacy_sources_loss": 0.7474144101142883, |
|
"eval_validation_privacy_sources_runtime": 190.0407, |
|
"eval_validation_privacy_sources_samples_per_second": 0.726, |
|
"eval_validation_privacy_sources_steps_per_second": 0.726, |
|
"step": 3553 |
|
}, |
|
{ |
|
"epoch": 0.75, |
|
"eval_validation_agenda_digitale_loss": 1.4130672216415405, |
|
"eval_validation_agenda_digitale_runtime": 170.6819, |
|
"eval_validation_agenda_digitale_samples_per_second": 0.726, |
|
"eval_validation_agenda_digitale_steps_per_second": 0.726, |
|
"step": 3553 |
|
}, |
|
{ |
|
"epoch": 0.75, |
|
"eval_validation_legal_articles_loss": 1.2912746667861938, |
|
"eval_validation_legal_articles_runtime": 273.9492, |
|
"eval_validation_legal_articles_samples_per_second": 0.726, |
|
"eval_validation_legal_articles_steps_per_second": 0.726, |
|
"step": 3553 |
|
}, |
|
{ |
|
"epoch": 0.75, |
|
"eval_validation_leggepertutti_loss": 1.1468725204467773, |
|
"eval_validation_leggepertutti_runtime": 26.1436, |
|
"eval_validation_leggepertutti_samples_per_second": 0.727, |
|
"eval_validation_leggepertutti_steps_per_second": 0.727, |
|
"step": 3553 |
|
}, |
|
{ |
|
"epoch": 0.78, |
|
"grad_norm": 2.265625, |
|
"learning_rate": 1e-05, |
|
"loss": 0.8363, |
|
"step": 3740 |
|
}, |
|
{ |
|
"epoch": 0.78, |
|
"eval_validation_loss": 0.7443498969078064, |
|
"eval_validation_runtime": 189.9118, |
|
"eval_validation_samples_per_second": 0.727, |
|
"eval_validation_steps_per_second": 0.727, |
|
"step": 3740 |
|
}, |
|
{ |
|
"epoch": 0.78, |
|
"eval_validation_privacy_sources_loss": 0.7443498969078064, |
|
"eval_validation_privacy_sources_runtime": 190.0301, |
|
"eval_validation_privacy_sources_samples_per_second": 0.726, |
|
"eval_validation_privacy_sources_steps_per_second": 0.726, |
|
"step": 3740 |
|
}, |
|
{ |
|
"epoch": 0.78, |
|
"eval_validation_agenda_digitale_loss": 1.4134020805358887, |
|
"eval_validation_agenda_digitale_runtime": 170.7516, |
|
"eval_validation_agenda_digitale_samples_per_second": 0.726, |
|
"eval_validation_agenda_digitale_steps_per_second": 0.726, |
|
"step": 3740 |
|
}, |
|
{ |
|
"epoch": 0.78, |
|
"eval_validation_legal_articles_loss": 1.2929922342300415, |
|
"eval_validation_legal_articles_runtime": 274.0196, |
|
"eval_validation_legal_articles_samples_per_second": 0.726, |
|
"eval_validation_legal_articles_steps_per_second": 0.726, |
|
"step": 3740 |
|
}, |
|
{ |
|
"epoch": 0.78, |
|
"eval_validation_leggepertutti_loss": 1.1504395008087158, |
|
"eval_validation_leggepertutti_runtime": 26.181, |
|
"eval_validation_leggepertutti_samples_per_second": 0.726, |
|
"eval_validation_leggepertutti_steps_per_second": 0.726, |
|
"step": 3740 |
|
}, |
|
{ |
|
"epoch": 0.82, |
|
"grad_norm": 2.125, |
|
"learning_rate": 1e-05, |
|
"loss": 0.8409, |
|
"step": 3927 |
|
}, |
|
{ |
|
"epoch": 0.82, |
|
"eval_validation_loss": 0.7407341599464417, |
|
"eval_validation_runtime": 190.0375, |
|
"eval_validation_samples_per_second": 0.726, |
|
"eval_validation_steps_per_second": 0.726, |
|
"step": 3927 |
|
}, |
|
{ |
|
"epoch": 0.82, |
|
"eval_validation_privacy_sources_loss": 0.7407341599464417, |
|
"eval_validation_privacy_sources_runtime": 190.0419, |
|
"eval_validation_privacy_sources_samples_per_second": 0.726, |
|
"eval_validation_privacy_sources_steps_per_second": 0.726, |
|
"step": 3927 |
|
}, |
|
{ |
|
"epoch": 0.82, |
|
"eval_validation_agenda_digitale_loss": 1.4093834161758423, |
|
"eval_validation_agenda_digitale_runtime": 170.7265, |
|
"eval_validation_agenda_digitale_samples_per_second": 0.726, |
|
"eval_validation_agenda_digitale_steps_per_second": 0.726, |
|
"step": 3927 |
|
}, |
|
{ |
|
"epoch": 0.82, |
|
"eval_validation_legal_articles_loss": 1.287365436553955, |
|
"eval_validation_legal_articles_runtime": 273.961, |
|
"eval_validation_legal_articles_samples_per_second": 0.726, |
|
"eval_validation_legal_articles_steps_per_second": 0.726, |
|
"step": 3927 |
|
}, |
|
{ |
|
"epoch": 0.82, |
|
"eval_validation_leggepertutti_loss": 1.1438554525375366, |
|
"eval_validation_leggepertutti_runtime": 26.1491, |
|
"eval_validation_leggepertutti_samples_per_second": 0.727, |
|
"eval_validation_leggepertutti_steps_per_second": 0.727, |
|
"step": 3927 |
|
}, |
|
{ |
|
"epoch": 0.86, |
|
"grad_norm": 2.3125, |
|
"learning_rate": 1e-05, |
|
"loss": 0.847, |
|
"step": 4114 |
|
}, |
|
{ |
|
"epoch": 0.86, |
|
"eval_validation_loss": 0.7385227680206299, |
|
"eval_validation_runtime": 190.0405, |
|
"eval_validation_samples_per_second": 0.726, |
|
"eval_validation_steps_per_second": 0.726, |
|
"step": 4114 |
|
}, |
|
{ |
|
"epoch": 0.86, |
|
"eval_validation_privacy_sources_loss": 0.7385227680206299, |
|
"eval_validation_privacy_sources_runtime": 190.0065, |
|
"eval_validation_privacy_sources_samples_per_second": 0.726, |
|
"eval_validation_privacy_sources_steps_per_second": 0.726, |
|
"step": 4114 |
|
}, |
|
{ |
|
"epoch": 0.86, |
|
"eval_validation_agenda_digitale_loss": 1.4071052074432373, |
|
"eval_validation_agenda_digitale_runtime": 170.7073, |
|
"eval_validation_agenda_digitale_samples_per_second": 0.726, |
|
"eval_validation_agenda_digitale_steps_per_second": 0.726, |
|
"step": 4114 |
|
}, |
|
{ |
|
"epoch": 0.86, |
|
"eval_validation_legal_articles_loss": 1.284498929977417, |
|
"eval_validation_legal_articles_runtime": 274.0307, |
|
"eval_validation_legal_articles_samples_per_second": 0.726, |
|
"eval_validation_legal_articles_steps_per_second": 0.726, |
|
"step": 4114 |
|
}, |
|
{ |
|
"epoch": 0.86, |
|
"eval_validation_leggepertutti_loss": 1.1409564018249512, |
|
"eval_validation_leggepertutti_runtime": 26.1558, |
|
"eval_validation_leggepertutti_samples_per_second": 0.726, |
|
"eval_validation_leggepertutti_steps_per_second": 0.726, |
|
"step": 4114 |
|
}, |
|
{ |
|
"epoch": 0.9, |
|
"grad_norm": 2.1875, |
|
"learning_rate": 1e-05, |
|
"loss": 0.8356, |
|
"step": 4301 |
|
}, |
|
{ |
|
"epoch": 0.9, |
|
"eval_validation_loss": 0.7365601658821106, |
|
"eval_validation_runtime": 190.0607, |
|
"eval_validation_samples_per_second": 0.726, |
|
"eval_validation_steps_per_second": 0.726, |
|
"step": 4301 |
|
}, |
|
{ |
|
"epoch": 0.9, |
|
"eval_validation_privacy_sources_loss": 0.7365601658821106, |
|
"eval_validation_privacy_sources_runtime": 190.1229, |
|
"eval_validation_privacy_sources_samples_per_second": 0.726, |
|
"eval_validation_privacy_sources_steps_per_second": 0.726, |
|
"step": 4301 |
|
}, |
|
{ |
|
"epoch": 0.9, |
|
"eval_validation_agenda_digitale_loss": 1.407541036605835, |
|
"eval_validation_agenda_digitale_runtime": 170.8311, |
|
"eval_validation_agenda_digitale_samples_per_second": 0.726, |
|
"eval_validation_agenda_digitale_steps_per_second": 0.726, |
|
"step": 4301 |
|
}, |
|
{ |
|
"epoch": 0.9, |
|
"eval_validation_legal_articles_loss": 1.2855533361434937, |
|
"eval_validation_legal_articles_runtime": 274.1776, |
|
"eval_validation_legal_articles_samples_per_second": 0.726, |
|
"eval_validation_legal_articles_steps_per_second": 0.726, |
|
"step": 4301 |
|
}, |
|
{ |
|
"epoch": 0.9, |
|
"eval_validation_leggepertutti_loss": 1.1440407037734985, |
|
"eval_validation_leggepertutti_runtime": 26.1983, |
|
"eval_validation_leggepertutti_samples_per_second": 0.725, |
|
"eval_validation_leggepertutti_steps_per_second": 0.725, |
|
"step": 4301 |
|
} |
|
], |
|
"logging_steps": 187, |
|
"max_steps": 4765, |
|
"num_input_tokens_seen": 0, |
|
"num_train_epochs": 1, |
|
"save_steps": 187, |
|
"total_flos": 4.8102819610591494e+19, |
|
"train_batch_size": 1, |
|
"trial_name": null, |
|
"trial_params": null |
|
} |
|
|