{ "best_metric": null, "best_model_checkpoint": null, "epoch": 0.5428675665815396, "eval_steps": 93, "global_step": 651, "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 } ], "logging_steps": 93, "max_steps": 2398, "num_input_tokens_seen": 0, "num_train_epochs": 2, "save_steps": 93, "total_flos": 7.305364071958708e+18, "train_batch_size": 1, "trial_name": null, "trial_params": null }