{ "best_metric": null, "best_model_checkpoint": null, "epoch": 20.0, "global_step": 300, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 1.67, "eval_accuracy": 0.6553743683968765, "eval_f1": 0.49752416882810657, "eval_loss": 1.2105501890182495, "eval_precision": 0.4707719767960732, "eval_recall": 0.5275, "eval_runtime": 3.73, "eval_samples_per_second": 14.477, "eval_steps_per_second": 1.609, "step": 25 }, { "epoch": 3.33, "eval_accuracy": 0.7687184198438217, "eval_f1": 0.777563793933558, "eval_loss": 0.7854474186897278, "eval_precision": 0.7497678737233054, "eval_recall": 0.8075, "eval_runtime": 3.7553, "eval_samples_per_second": 14.38, "eval_steps_per_second": 1.598, "step": 50 }, { "epoch": 5.0, "eval_accuracy": 0.8141938447404685, "eval_f1": 0.8184898354307841, "eval_loss": 0.6001709699630737, "eval_precision": 0.7931519699812383, "eval_recall": 0.8455, "eval_runtime": 3.7036, "eval_samples_per_second": 14.58, "eval_steps_per_second": 1.62, "step": 75 }, { "epoch": 6.67, "eval_accuracy": 0.7781350482315113, "eval_f1": 0.827977315689981, "eval_loss": 0.6523196697235107, "eval_precision": 0.7849462365591398, "eval_recall": 0.876, "eval_runtime": 3.7641, "eval_samples_per_second": 14.346, "eval_steps_per_second": 1.594, "step": 100 }, { "epoch": 8.33, "eval_accuracy": 0.8354386770785485, "eval_f1": 0.8442622950819673, "eval_loss": 0.5189960598945618, "eval_precision": 0.8151769087523277, "eval_recall": 0.8755, "eval_runtime": 3.8049, "eval_samples_per_second": 14.192, "eval_steps_per_second": 1.577, "step": 125 }, { "epoch": 10.0, "eval_accuracy": 0.8338309600367478, "eval_f1": 0.8588007736943907, "eval_loss": 0.5064252018928528, "eval_precision": 0.8314606741573034, "eval_recall": 0.888, "eval_runtime": 3.7776, "eval_samples_per_second": 14.295, "eval_steps_per_second": 1.588, "step": 150 }, { "epoch": 11.67, "eval_accuracy": 0.8344051446945338, "eval_f1": 0.8693320331545589, "eval_loss": 0.5342020392417908, "eval_precision": 0.8482397716460514, "eval_recall": 0.8915, "eval_runtime": 3.753, "eval_samples_per_second": 14.388, "eval_steps_per_second": 1.599, "step": 175 }, { "epoch": 13.33, "eval_accuracy": 0.8200505282498851, "eval_f1": 0.8703071672354948, "eval_loss": 0.5538159012794495, "eval_precision": 0.8491912464319695, "eval_recall": 0.8925, "eval_runtime": 3.8395, "eval_samples_per_second": 14.064, "eval_steps_per_second": 1.563, "step": 200 }, { "epoch": 15.0, "eval_accuracy": 0.8348644924207625, "eval_f1": 0.8777398928397467, "eval_loss": 0.5335590243339539, "eval_precision": 0.855650522317189, "eval_recall": 0.901, "eval_runtime": 3.7278, "eval_samples_per_second": 14.486, "eval_steps_per_second": 1.61, "step": 225 }, { "epoch": 16.67, "eval_accuracy": 0.8385392742305926, "eval_f1": 0.8764648437499999, "eval_loss": 0.5464726686477661, "eval_precision": 0.8563931297709924, "eval_recall": 0.8975, "eval_runtime": 3.7679, "eval_samples_per_second": 14.331, "eval_steps_per_second": 1.592, "step": 250 }, { "epoch": 18.33, "eval_accuracy": 0.8439366100137804, "eval_f1": 0.8787509148572822, "eval_loss": 0.5402917265892029, "eval_precision": 0.8580276322058122, "eval_recall": 0.9005, "eval_runtime": 3.7992, "eval_samples_per_second": 14.214, "eval_steps_per_second": 1.579, "step": 275 }, { "epoch": 20.0, "eval_accuracy": 0.8434772622875517, "eval_f1": 0.8826979472140762, "eval_loss": 0.5468625426292419, "eval_precision": 0.8632887189292543, "eval_recall": 0.903, "eval_runtime": 3.7775, "eval_samples_per_second": 14.295, "eval_steps_per_second": 1.588, "step": 300 }, { "epoch": 20.0, "step": 300, "total_flos": 790816029696000.0, "train_loss": 0.4636675516764323, "train_runtime": 318.7136, "train_samples_per_second": 9.413, "train_steps_per_second": 0.941 } ], "max_steps": 300, "num_train_epochs": 20, "total_flos": 790816029696000.0, "trial_name": null, "trial_params": null }