ProMeText/aquilign-multilingual-segmenter

1{2  "best_metric": 0.014524628408253193,3  "best_model_checkpoint": "/projects/users/mgillele/trash/test_segmenter/2025-10-19T11:53:45.914349_entra\u00eenement_a0f1a64_v2/models/checkpoint-498",4  "epoch": 2.0,5  "eval_steps": 500,6  "global_step": 498,7  "is_hyper_param_search": false,8  "is_local_process_zero": true,9  "is_world_process_zero": true,10  "log_history": [11    {12      "epoch": 1.0,13      "grad_norm": 0.08521145582199097,14      "learning_rate": 4.8333333333333334e-05,15      "loss": 0.0363,16      "step": 24917    },18    {19      "epoch": 1.0,20      "eval_accuracy": {21        "accuracy": 0.973590718587804222      },23      "eval_f1": [24        0.9852987755417358,25        0.87028832390430126      ],27      "eval_loss": 0.01460354682058096,28      "eval_precision": [29        0.9857633092190689,30        0.866684767852294331      ],32      "eval_recall": [33        0.98483467947431,34        0.873921971252566835      ],36      "eval_runtime": 4.2136,37      "eval_samples_per_second": 236.138,38      "eval_steps_per_second": 3.797,39      "step": 24940    },41    {42      "epoch": 2.0,43      "grad_norm": 0.0830661952495575,44      "learning_rate": 4.666666666666667e-05,45      "loss": 0.012,46      "step": 49847    },48    {49      "epoch": 2.0,50      "eval_accuracy": {51        "accuracy": 0.974326237197812952      },53      "eval_f1": [54        0.9857028037960988,55        0.874320652173913156      ],57      "eval_loss": 0.014524628408253193,58      "eval_precision": [59        0.9865414662067045,60        0.867835468644639261      ],62      "eval_recall": [63        0.9848655660741588,64        0.880903490759753665      ],66      "eval_runtime": 4.2542,67      "eval_samples_per_second": 233.886,68      "eval_steps_per_second": 3.761,69      "step": 49870    }71  ],72  "logging_steps": 500,73  "max_steps": 7470,74  "num_input_tokens_seen": 0,75  "num_train_epochs": 30,76  "save_steps": 500,77  "stateful_callbacks": {78    "EarlyStoppingCallback": {79      "args": {80        "early_stopping_patience": 10,81        "early_stopping_threshold": 0.082      },83      "attributes": {84        "early_stopping_patience_counter": 085      }86    },87    "TrainerControl": {88      "args": {89        "should_epoch_stop": false,90        "should_evaluate": false,91        "should_log": false,92        "should_save": true,93        "should_training_stop": false94      },95      "attributes": {}96    }97  },98  "total_flos": 5721993434919168.0,99  "train_batch_size": 64,100  "trial_name": null,101  "trial_params": null102}103