{ "best_metric": null, "best_model_checkpoint": null, "epoch": 0.23360399750156152, "eval_steps": 187, "global_step": 935, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.00024984384759525296, "grad_norm": 72.0, "learning_rate": 2.5000000000000004e-07, "loss": 2.0834, "step": 1 }, { "epoch": 0.046720799500312306, "grad_norm": 0.58984375, "learning_rate": 1e-05, "loss": 1.51, "step": 187 }, { "epoch": 0.046720799500312306, "eval_validation_loss": 1.2587465047836304, "eval_validation_runtime": 107.1911, "eval_validation_samples_per_second": 1.119, "eval_validation_steps_per_second": 1.119, "step": 187 }, { "epoch": 0.046720799500312306, "eval_validation_privacy_sources_loss": 1.2587465047836304, "eval_validation_privacy_sources_runtime": 107.266, "eval_validation_privacy_sources_samples_per_second": 1.119, "eval_validation_privacy_sources_steps_per_second": 1.119, "step": 187 }, { "epoch": 0.046720799500312306, "eval_validation_agenda_digitale_loss": 1.7502987384796143, "eval_validation_agenda_digitale_runtime": 97.3699, "eval_validation_agenda_digitale_samples_per_second": 1.119, "eval_validation_agenda_digitale_steps_per_second": 1.119, "step": 187 }, { "epoch": 0.046720799500312306, "eval_validation_legal_articles_loss": 1.6148390769958496, "eval_validation_legal_articles_runtime": 155.4352, "eval_validation_legal_articles_samples_per_second": 1.119, "eval_validation_legal_articles_steps_per_second": 1.119, "step": 187 }, { "epoch": 0.046720799500312306, "eval_validation_leggepertutti_loss": 1.400872826576233, "eval_validation_leggepertutti_runtime": 15.1917, "eval_validation_leggepertutti_samples_per_second": 1.119, "eval_validation_leggepertutti_steps_per_second": 1.119, "step": 187 }, { "epoch": 0.09344159900062461, "grad_norm": 0.58984375, "learning_rate": 1e-05, "loss": 1.4038, "step": 374 }, { "epoch": 0.09344159900062461, "eval_validation_loss": 1.2273409366607666, "eval_validation_runtime": 107.1014, "eval_validation_samples_per_second": 1.12, "eval_validation_steps_per_second": 1.12, "step": 374 }, { "epoch": 0.09344159900062461, "eval_validation_privacy_sources_loss": 1.2273409366607666, "eval_validation_privacy_sources_runtime": 107.0462, "eval_validation_privacy_sources_samples_per_second": 1.121, "eval_validation_privacy_sources_steps_per_second": 1.121, "step": 374 }, { "epoch": 0.09344159900062461, "eval_validation_agenda_digitale_loss": 1.745036005973816, "eval_validation_agenda_digitale_runtime": 97.2787, "eval_validation_agenda_digitale_samples_per_second": 1.12, "eval_validation_agenda_digitale_steps_per_second": 1.12, "step": 374 }, { "epoch": 0.09344159900062461, "eval_validation_legal_articles_loss": 1.6085145473480225, "eval_validation_legal_articles_runtime": 155.2206, "eval_validation_legal_articles_samples_per_second": 1.121, "eval_validation_legal_articles_steps_per_second": 1.121, "step": 374 }, { "epoch": 0.09344159900062461, "eval_validation_leggepertutti_loss": 1.3941675424575806, "eval_validation_leggepertutti_runtime": 15.1745, "eval_validation_leggepertutti_samples_per_second": 1.12, "eval_validation_leggepertutti_steps_per_second": 1.12, "step": 374 }, { "epoch": 0.14016239850093692, "grad_norm": 0.53125, "learning_rate": 1e-05, "loss": 1.3853, "step": 561 }, { "epoch": 0.14016239850093692, "eval_validation_loss": 1.2080044746398926, "eval_validation_runtime": 107.2381, "eval_validation_samples_per_second": 1.119, "eval_validation_steps_per_second": 1.119, "step": 561 }, { "epoch": 0.14016239850093692, "eval_validation_privacy_sources_loss": 1.2080044746398926, "eval_validation_privacy_sources_runtime": 107.2154, "eval_validation_privacy_sources_samples_per_second": 1.119, "eval_validation_privacy_sources_steps_per_second": 1.119, "step": 561 }, { "epoch": 0.14016239850093692, "eval_validation_agenda_digitale_loss": 1.7419259548187256, "eval_validation_agenda_digitale_runtime": 97.4488, "eval_validation_agenda_digitale_samples_per_second": 1.119, "eval_validation_agenda_digitale_steps_per_second": 1.119, "step": 561 }, { "epoch": 0.14016239850093692, "eval_validation_legal_articles_loss": 1.6046104431152344, "eval_validation_legal_articles_runtime": 155.5238, "eval_validation_legal_articles_samples_per_second": 1.119, "eval_validation_legal_articles_steps_per_second": 1.119, "step": 561 }, { "epoch": 0.14016239850093692, "eval_validation_leggepertutti_loss": 1.3902024030685425, "eval_validation_leggepertutti_runtime": 15.1931, "eval_validation_leggepertutti_samples_per_second": 1.119, "eval_validation_leggepertutti_steps_per_second": 1.119, "step": 561 }, { "epoch": 0.18688319800124922, "grad_norm": 0.58984375, "learning_rate": 1e-05, "loss": 1.3639, "step": 748 }, { "epoch": 0.18688319800124922, "eval_validation_loss": 1.1931297779083252, "eval_validation_runtime": 107.2136, "eval_validation_samples_per_second": 1.119, "eval_validation_steps_per_second": 1.119, "step": 748 }, { "epoch": 0.18688319800124922, "eval_validation_privacy_sources_loss": 1.1931297779083252, "eval_validation_privacy_sources_runtime": 107.2505, "eval_validation_privacy_sources_samples_per_second": 1.119, "eval_validation_privacy_sources_steps_per_second": 1.119, "step": 748 }, { "epoch": 0.18688319800124922, "eval_validation_agenda_digitale_loss": 1.739527940750122, "eval_validation_agenda_digitale_runtime": 97.3978, "eval_validation_agenda_digitale_samples_per_second": 1.119, "eval_validation_agenda_digitale_steps_per_second": 1.119, "step": 748 }, { "epoch": 0.18688319800124922, "eval_validation_legal_articles_loss": 1.601611614227295, "eval_validation_legal_articles_runtime": 155.4821, "eval_validation_legal_articles_samples_per_second": 1.119, "eval_validation_legal_articles_steps_per_second": 1.119, "step": 748 }, { "epoch": 0.18688319800124922, "eval_validation_leggepertutti_loss": 1.3867729902267456, "eval_validation_leggepertutti_runtime": 15.1827, "eval_validation_leggepertutti_samples_per_second": 1.12, "eval_validation_leggepertutti_steps_per_second": 1.12, "step": 748 }, { "epoch": 0.23360399750156152, "grad_norm": 0.58203125, "learning_rate": 1e-05, "loss": 1.3565, "step": 935 }, { "epoch": 0.23360399750156152, "eval_validation_loss": 1.182249903678894, "eval_validation_runtime": 107.2517, "eval_validation_samples_per_second": 1.119, "eval_validation_steps_per_second": 1.119, "step": 935 }, { "epoch": 0.23360399750156152, "eval_validation_privacy_sources_loss": 1.182249903678894, "eval_validation_privacy_sources_runtime": 107.2175, "eval_validation_privacy_sources_samples_per_second": 1.119, "eval_validation_privacy_sources_steps_per_second": 1.119, "step": 935 }, { "epoch": 0.23360399750156152, "eval_validation_agenda_digitale_loss": 1.7376903295516968, "eval_validation_agenda_digitale_runtime": 97.43, "eval_validation_agenda_digitale_samples_per_second": 1.119, "eval_validation_agenda_digitale_steps_per_second": 1.119, "step": 935 }, { "epoch": 0.23360399750156152, "eval_validation_legal_articles_loss": 1.5993213653564453, "eval_validation_legal_articles_runtime": 155.5081, "eval_validation_legal_articles_samples_per_second": 1.119, "eval_validation_legal_articles_steps_per_second": 1.119, "step": 935 }, { "epoch": 0.23360399750156152, "eval_validation_leggepertutti_loss": 1.3844820261001587, "eval_validation_leggepertutti_runtime": 15.1978, "eval_validation_leggepertutti_samples_per_second": 1.119, "eval_validation_leggepertutti_steps_per_second": 1.119, "step": 935 } ], "logging_steps": 187, "max_steps": 4002, "num_input_tokens_seen": 0, "num_train_epochs": 1, "save_steps": 187, "stateful_callbacks": { "TrainerControl": { "args": { "should_epoch_stop": false, "should_evaluate": false, "should_log": false, "should_save": true, "should_training_stop": false }, "attributes": {} } }, "total_flos": 4.1453283886025933e+18, "train_batch_size": 1, "trial_name": null, "trial_params": null }