|
{ |
|
"best_metric": 0.6690246389304302, |
|
"best_model_checkpoint": "case-analysis-InLegalBERT/checkpoint-1344", |
|
"epoch": 11.0, |
|
"eval_steps": 500, |
|
"global_step": 2464, |
|
"is_hyper_param_search": false, |
|
"is_local_process_zero": true, |
|
"is_world_process_zero": true, |
|
"log_history": [ |
|
{ |
|
"epoch": 1.0, |
|
"eval_accuracy": 0.8017817371937639, |
|
"eval_f1_macro": 0.5936102522033613, |
|
"eval_f1_micro": 0.8017817371937639, |
|
"eval_f1_weighted": 0.7819923339877444, |
|
"eval_loss": 0.6545999050140381, |
|
"eval_macro_fpr": 0.09781883735656886, |
|
"eval_macro_sensitivity": 0.6106003826592061, |
|
"eval_macro_specificity": 0.9112420642353884, |
|
"eval_precision": 0.763240968439436, |
|
"eval_precision_macro": 0.5776616905319997, |
|
"eval_recall": 0.8017817371937639, |
|
"eval_recall_macro": 0.6106003826592061, |
|
"eval_runtime": 9.2925, |
|
"eval_samples_per_second": 48.318, |
|
"eval_steps_per_second": 6.134, |
|
"eval_weighted_fpr": 0.07613344739093242, |
|
"eval_weighted_sensitivity": 0.8017817371937639, |
|
"eval_weighted_specificity": 0.84318651974779, |
|
"step": 224 |
|
}, |
|
{ |
|
"epoch": 2.0, |
|
"eval_accuracy": 0.8129175946547884, |
|
"eval_f1_macro": 0.5995597405020452, |
|
"eval_f1_micro": 0.8129175946547884, |
|
"eval_f1_weighted": 0.7925821933419933, |
|
"eval_loss": 0.6830806732177734, |
|
"eval_macro_fpr": 0.0922965769071398, |
|
"eval_macro_sensitivity": 0.615426637485461, |
|
"eval_macro_specificity": 0.9170902440637451, |
|
"eval_precision": 0.7732439444566204, |
|
"eval_precision_macro": 0.5844969763606377, |
|
"eval_recall": 0.8129175946547884, |
|
"eval_recall_macro": 0.615426637485461, |
|
"eval_runtime": 36.5377, |
|
"eval_samples_per_second": 12.289, |
|
"eval_steps_per_second": 1.56, |
|
"eval_weighted_fpr": 0.07124681933842239, |
|
"eval_weighted_sensitivity": 0.8129175946547884, |
|
"eval_weighted_specificity": 0.8554433816001921, |
|
"step": 448 |
|
}, |
|
{ |
|
"epoch": 2.23, |
|
"grad_norm": 6.910687446594238, |
|
"learning_rate": 4.6302083333333335e-05, |
|
"loss": 0.607, |
|
"step": 500 |
|
}, |
|
{ |
|
"epoch": 3.0, |
|
"eval_accuracy": 0.8262806236080178, |
|
"eval_f1_macro": 0.6362135046138226, |
|
"eval_f1_micro": 0.8262806236080178, |
|
"eval_f1_weighted": 0.8105096248986687, |
|
"eval_loss": 0.7626442909240723, |
|
"eval_macro_fpr": 0.0885057964533289, |
|
"eval_macro_sensitivity": 0.6341414944356121, |
|
"eval_macro_specificity": 0.9181782637863818, |
|
"eval_precision": 0.8060163824047696, |
|
"eval_precision_macro": 0.6772816989326722, |
|
"eval_recall": 0.8262806236080178, |
|
"eval_recall_macro": 0.6341414944356121, |
|
"eval_runtime": 35.6726, |
|
"eval_samples_per_second": 12.587, |
|
"eval_steps_per_second": 1.598, |
|
"eval_weighted_fpr": 0.0654911838790932, |
|
"eval_weighted_sensitivity": 0.8262806236080178, |
|
"eval_weighted_specificity": 0.8464324315375095, |
|
"step": 672 |
|
}, |
|
{ |
|
"epoch": 4.0, |
|
"eval_accuracy": 0.8084632516703786, |
|
"eval_f1_macro": 0.6313839669673806, |
|
"eval_f1_micro": 0.8084632516703786, |
|
"eval_f1_weighted": 0.8017481865377005, |
|
"eval_loss": 0.7839008569717407, |
|
"eval_macro_fpr": 0.08961602877899306, |
|
"eval_macro_sensitivity": 0.630579873226932, |
|
"eval_macro_specificity": 0.9209618973154744, |
|
"eval_precision": 0.7990879202460494, |
|
"eval_precision_macro": 0.6390850934968582, |
|
"eval_recall": 0.8084632516703786, |
|
"eval_recall_macro": 0.630579873226932, |
|
"eval_runtime": 33.6763, |
|
"eval_samples_per_second": 13.333, |
|
"eval_steps_per_second": 1.693, |
|
"eval_weighted_fpr": 0.07319148936170213, |
|
"eval_weighted_sensitivity": 0.8084632516703786, |
|
"eval_weighted_specificity": 0.8753843375915193, |
|
"step": 896 |
|
}, |
|
{ |
|
"epoch": 4.46, |
|
"grad_norm": 0.3201627731323242, |
|
"learning_rate": 4.258184523809524e-05, |
|
"loss": 0.316, |
|
"step": 1000 |
|
}, |
|
{ |
|
"epoch": 5.0, |
|
"eval_accuracy": 0.8262806236080178, |
|
"eval_f1_macro": 0.6513945578231293, |
|
"eval_f1_micro": 0.8262806236080178, |
|
"eval_f1_weighted": 0.8161065406117904, |
|
"eval_loss": 0.9380576610565186, |
|
"eval_macro_fpr": 0.08224137352531583, |
|
"eval_macro_sensitivity": 0.6573161205514146, |
|
"eval_macro_specificity": 0.9260657452373569, |
|
"eval_precision": 0.8127349298285333, |
|
"eval_precision_macro": 0.668810150869688, |
|
"eval_recall": 0.8262806236080178, |
|
"eval_recall_macro": 0.6573161205514146, |
|
"eval_runtime": 34.3275, |
|
"eval_samples_per_second": 13.08, |
|
"eval_steps_per_second": 1.66, |
|
"eval_weighted_fpr": 0.0654911838790932, |
|
"eval_weighted_sensitivity": 0.8262806236080178, |
|
"eval_weighted_specificity": 0.8779823573414097, |
|
"step": 1120 |
|
}, |
|
{ |
|
"epoch": 6.0, |
|
"eval_accuracy": 0.821826280623608, |
|
"eval_f1_macro": 0.6690246389304302, |
|
"eval_f1_micro": 0.821826280623608, |
|
"eval_f1_weighted": 0.8159237291984361, |
|
"eval_loss": 1.0433590412139893, |
|
"eval_macro_fpr": 0.08967976675610558, |
|
"eval_macro_sensitivity": 0.6533037047742931, |
|
"eval_macro_specificity": 0.9186687903009794, |
|
"eval_precision": 0.8145202780918083, |
|
"eval_precision_macro": 0.6906556062141168, |
|
"eval_recall": 0.821826280623608, |
|
"eval_recall_macro": 0.6533037047742931, |
|
"eval_runtime": 43.316, |
|
"eval_samples_per_second": 10.366, |
|
"eval_steps_per_second": 1.316, |
|
"eval_weighted_fpr": 0.06739679865206402, |
|
"eval_weighted_sensitivity": 0.821826280623608, |
|
"eval_weighted_specificity": 0.8528488805803094, |
|
"step": 1344 |
|
}, |
|
{ |
|
"epoch": 6.7, |
|
"grad_norm": 0.028867056593298912, |
|
"learning_rate": 3.8869047619047625e-05, |
|
"loss": 0.1513, |
|
"step": 1500 |
|
}, |
|
{ |
|
"epoch": 7.0, |
|
"eval_accuracy": 0.8017817371937639, |
|
"eval_f1_macro": 0.6374820990180032, |
|
"eval_f1_micro": 0.8017817371937639, |
|
"eval_f1_weighted": 0.8030061499337195, |
|
"eval_loss": 1.2181551456451416, |
|
"eval_macro_fpr": 0.09161393040768498, |
|
"eval_macro_sensitivity": 0.6398633163339046, |
|
"eval_macro_specificity": 0.9204902964282863, |
|
"eval_precision": 0.8065681635796801, |
|
"eval_precision_macro": 0.6382217224720973, |
|
"eval_recall": 0.8017817371937639, |
|
"eval_recall_macro": 0.6398633163339046, |
|
"eval_runtime": 38.5257, |
|
"eval_samples_per_second": 11.655, |
|
"eval_steps_per_second": 1.48, |
|
"eval_weighted_fpr": 0.07613344739093242, |
|
"eval_weighted_sensitivity": 0.8017817371937639, |
|
"eval_weighted_specificity": 0.8801794485193812, |
|
"step": 1568 |
|
}, |
|
{ |
|
"epoch": 8.0, |
|
"eval_accuracy": 0.8285077951002228, |
|
"eval_f1_macro": 0.6375618771468077, |
|
"eval_f1_micro": 0.8285077951002228, |
|
"eval_f1_weighted": 0.8152450374079808, |
|
"eval_loss": 1.3192639350891113, |
|
"eval_macro_fpr": 0.08815434538353528, |
|
"eval_macro_sensitivity": 0.6279757603287015, |
|
"eval_macro_specificity": 0.9201548104191046, |
|
"eval_precision": 0.8069973553326301, |
|
"eval_precision_macro": 0.6565998635009741, |
|
"eval_recall": 0.8285077951002228, |
|
"eval_recall_macro": 0.6279757603287015, |
|
"eval_runtime": 33.4932, |
|
"eval_samples_per_second": 13.406, |
|
"eval_steps_per_second": 1.702, |
|
"eval_weighted_fpr": 0.06454316848281642, |
|
"eval_weighted_sensitivity": 0.8285077951002228, |
|
"eval_weighted_specificity": 0.8521114465761961, |
|
"step": 1792 |
|
}, |
|
{ |
|
"epoch": 8.93, |
|
"grad_norm": 0.008128850720822811, |
|
"learning_rate": 3.514880952380952e-05, |
|
"loss": 0.0491, |
|
"step": 2000 |
|
}, |
|
{ |
|
"epoch": 9.0, |
|
"eval_accuracy": 0.8329621380846325, |
|
"eval_f1_macro": 0.6687059701183248, |
|
"eval_f1_micro": 0.8329621380846325, |
|
"eval_f1_weighted": 0.8234691462881575, |
|
"eval_loss": 1.3168540000915527, |
|
"eval_macro_fpr": 0.08276943574584078, |
|
"eval_macro_sensitivity": 0.6554578833990599, |
|
"eval_macro_specificity": 0.9245723745022232, |
|
"eval_precision": 0.8179557632761028, |
|
"eval_precision_macro": 0.6949834258114409, |
|
"eval_recall": 0.8329621380846325, |
|
"eval_recall_macro": 0.6554578833990599, |
|
"eval_runtime": 36.7708, |
|
"eval_samples_per_second": 12.211, |
|
"eval_steps_per_second": 1.55, |
|
"eval_weighted_fpr": 0.06265664160401002, |
|
"eval_weighted_sensitivity": 0.8329621380846325, |
|
"eval_weighted_specificity": 0.86532735992426, |
|
"step": 2016 |
|
}, |
|
{ |
|
"epoch": 10.0, |
|
"eval_accuracy": 0.8307349665924276, |
|
"eval_f1_macro": 0.6397801007606054, |
|
"eval_f1_micro": 0.8307349665924276, |
|
"eval_f1_weighted": 0.8183915457597033, |
|
"eval_loss": 1.4459574222564697, |
|
"eval_macro_fpr": 0.08684003094467867, |
|
"eval_macro_sensitivity": 0.6290898423251364, |
|
"eval_macro_specificity": 0.9210102525156295, |
|
"eval_precision": 0.8108965608714154, |
|
"eval_precision_macro": 0.6584470636889992, |
|
"eval_recall": 0.8307349665924276, |
|
"eval_recall_macro": 0.6290898423251364, |
|
"eval_runtime": 36.4367, |
|
"eval_samples_per_second": 12.323, |
|
"eval_steps_per_second": 1.564, |
|
"eval_weighted_fpr": 0.06359832635983263, |
|
"eval_weighted_sensitivity": 0.8307349665924276, |
|
"eval_weighted_specificity": 0.8533060434700906, |
|
"step": 2240 |
|
}, |
|
{ |
|
"epoch": 11.0, |
|
"eval_accuracy": 0.8418708240534521, |
|
"eval_f1_macro": 0.6463830564509832, |
|
"eval_f1_micro": 0.8418708240534521, |
|
"eval_f1_weighted": 0.8263008274936262, |
|
"eval_loss": 1.4100018739700317, |
|
"eval_macro_fpr": 0.08061341387649999, |
|
"eval_macro_sensitivity": 0.6398590148590149, |
|
"eval_macro_specificity": 0.9265216670805116, |
|
"eval_precision": 0.8165947524130448, |
|
"eval_precision_macro": 0.6717971143983426, |
|
"eval_recall": 0.8418708240534521, |
|
"eval_recall_macro": 0.6398590148590149, |
|
"eval_runtime": 33.4877, |
|
"eval_samples_per_second": 13.408, |
|
"eval_steps_per_second": 1.702, |
|
"eval_weighted_fpr": 0.05892116182572614, |
|
"eval_weighted_sensitivity": 0.8418708240534521, |
|
"eval_weighted_specificity": 0.8642158442685944, |
|
"step": 2464 |
|
} |
|
], |
|
"logging_steps": 500, |
|
"max_steps": 6720, |
|
"num_input_tokens_seen": 0, |
|
"num_train_epochs": 30, |
|
"save_steps": 500, |
|
"total_flos": 5186538257252352.0, |
|
"train_batch_size": 8, |
|
"trial_name": null, |
|
"trial_params": null |
|
} |
|
|