| { |
| "best_metric": 0.9915220559014439, |
| "best_model_checkpoint": ".//debugged_cz_cac_ses_udpipe_16_0.01_0.00002_20_04-23-24_04-34/checkpoint-20995", |
| "epoch": 19.999595305544315, |
| "global_step": 24700, |
| "is_hyper_param_search": false, |
| "is_local_process_zero": true, |
| "is_world_process_zero": true, |
| "log_history": [ |
| { |
| "epoch": 1.0, |
| "learning_rate": 1.6666666666666667e-05, |
| "loss": 1.6616, |
| "step": 1235 |
| }, |
| { |
| "epoch": 1.0, |
| "eval_accuracy": 0.9560764270169905, |
| "eval_f1": 0.9400045099417687, |
| "eval_loss": 0.24223673343658447, |
| "eval_precision": 0.9412405365918448, |
| "eval_recall": 0.9387717253073337, |
| "eval_runtime": 27.0992, |
| "eval_samples_per_second": 91.221, |
| "eval_steps_per_second": 11.403, |
| "step": 1235 |
| }, |
| { |
| "epoch": 2.0, |
| "learning_rate": 1.914893617021277e-05, |
| "loss": 0.1798, |
| "step": 2470 |
| }, |
| { |
| "epoch": 2.0, |
| "eval_accuracy": 0.9818715435142852, |
| "eval_f1": 0.9734886124249771, |
| "eval_loss": 0.09523347020149231, |
| "eval_precision": 0.9736305091034373, |
| "eval_recall": 0.9733467571004663, |
| "eval_runtime": 22.0988, |
| "eval_samples_per_second": 111.861, |
| "eval_steps_per_second": 13.983, |
| "step": 2470 |
| }, |
| { |
| "epoch": 3.0, |
| "learning_rate": 1.8085106382978724e-05, |
| "loss": 0.0815, |
| "step": 3705 |
| }, |
| { |
| "epoch": 3.0, |
| "eval_accuracy": 0.987861121648332, |
| "eval_f1": 0.9822315856421673, |
| "eval_loss": 0.06692199409008026, |
| "eval_precision": 0.9824528852015798, |
| "eval_recall": 0.9820103857566765, |
| "eval_runtime": 22.0277, |
| "eval_samples_per_second": 112.223, |
| "eval_steps_per_second": 14.028, |
| "step": 3705 |
| }, |
| { |
| "epoch": 4.0, |
| "learning_rate": 1.7021276595744682e-05, |
| "loss": 0.0475, |
| "step": 4940 |
| }, |
| { |
| "epoch": 4.0, |
| "eval_accuracy": 0.9899974045161419, |
| "eval_f1": 0.9855729104566591, |
| "eval_loss": 0.05793284624814987, |
| "eval_precision": 0.9856381992103659, |
| "eval_recall": 0.985507630351844, |
| "eval_runtime": 21.9964, |
| "eval_samples_per_second": 112.382, |
| "eval_steps_per_second": 14.048, |
| "step": 4940 |
| }, |
| { |
| "epoch": 5.0, |
| "learning_rate": 1.595744680851064e-05, |
| "loss": 0.0339, |
| "step": 6175 |
| }, |
| { |
| "epoch": 5.0, |
| "eval_accuracy": 0.9916145906123346, |
| "eval_f1": 0.9877177873468036, |
| "eval_loss": 0.051405712962150574, |
| "eval_precision": 0.9878879436007527, |
| "eval_recall": 0.987547689699025, |
| "eval_runtime": 22.0233, |
| "eval_samples_per_second": 112.245, |
| "eval_steps_per_second": 14.031, |
| "step": 6175 |
| }, |
| { |
| "epoch": 6.0, |
| "learning_rate": 1.4893617021276596e-05, |
| "loss": 0.0235, |
| "step": 7410 |
| }, |
| { |
| "epoch": 6.0, |
| "eval_accuracy": 0.9920737916026114, |
| "eval_f1": 0.9885550786838342, |
| "eval_loss": 0.04887189716100693, |
| "eval_precision": 0.9885027021299142, |
| "eval_recall": 0.9886074607884697, |
| "eval_runtime": 22.0452, |
| "eval_samples_per_second": 112.133, |
| "eval_steps_per_second": 14.017, |
| "step": 7410 |
| }, |
| { |
| "epoch": 7.0, |
| "learning_rate": 1.3829787234042556e-05, |
| "loss": 0.0172, |
| "step": 8645 |
| }, |
| { |
| "epoch": 7.0, |
| "eval_accuracy": 0.9921336873839519, |
| "eval_f1": 0.9884765159341969, |
| "eval_loss": 0.052920494228601456, |
| "eval_precision": 0.9883456057636277, |
| "eval_recall": 0.9886074607884697, |
| "eval_runtime": 22.0116, |
| "eval_samples_per_second": 112.304, |
| "eval_steps_per_second": 14.038, |
| "step": 8645 |
| }, |
| { |
| "epoch": 8.0, |
| "learning_rate": 1.2765957446808513e-05, |
| "loss": 0.0135, |
| "step": 9880 |
| }, |
| { |
| "epoch": 8.0, |
| "eval_accuracy": 0.9927725757182503, |
| "eval_f1": 0.9894034121013034, |
| "eval_loss": 0.047851815819740295, |
| "eval_precision": 0.9892985802076711, |
| "eval_recall": 0.9895082662144977, |
| "eval_runtime": 22.0657, |
| "eval_samples_per_second": 112.029, |
| "eval_steps_per_second": 14.004, |
| "step": 9880 |
| }, |
| { |
| "epoch": 9.0, |
| "learning_rate": 1.170212765957447e-05, |
| "loss": 0.0104, |
| "step": 11115 |
| }, |
| { |
| "epoch": 9.0, |
| "eval_accuracy": 0.992572923113782, |
| "eval_f1": 0.9889149350391355, |
| "eval_loss": 0.051443468779325485, |
| "eval_precision": 0.9886661547016922, |
| "eval_recall": 0.9891638406104282, |
| "eval_runtime": 21.9418, |
| "eval_samples_per_second": 112.662, |
| "eval_steps_per_second": 14.083, |
| "step": 11115 |
| }, |
| { |
| "epoch": 10.0, |
| "learning_rate": 1.0638297872340426e-05, |
| "loss": 0.0077, |
| "step": 12350 |
| }, |
| { |
| "epoch": 10.0, |
| "eval_accuracy": 0.9928724020204843, |
| "eval_f1": 0.9894731266799963, |
| "eval_loss": 0.04969193413853645, |
| "eval_precision": 0.9890409508431056, |
| "eval_recall": 0.9899056803730394, |
| "eval_runtime": 22.1653, |
| "eval_samples_per_second": 111.526, |
| "eval_steps_per_second": 13.941, |
| "step": 12350 |
| }, |
| { |
| "epoch": 11.0, |
| "learning_rate": 9.574468085106385e-06, |
| "loss": 0.006, |
| "step": 13585 |
| }, |
| { |
| "epoch": 11.0, |
| "eval_accuracy": 0.992912332541378, |
| "eval_f1": 0.9894125909337856, |
| "eval_loss": 0.05363365635275841, |
| "eval_precision": 0.9896879887601728, |
| "eval_recall": 0.9891373463331921, |
| "eval_runtime": 22.4359, |
| "eval_samples_per_second": 110.181, |
| "eval_steps_per_second": 13.773, |
| "step": 13585 |
| }, |
| { |
| "epoch": 12.0, |
| "learning_rate": 8.510638297872341e-06, |
| "loss": 0.005, |
| "step": 14820 |
| }, |
| { |
| "epoch": 12.0, |
| "eval_accuracy": 0.9936909776988041, |
| "eval_f1": 0.9908487729939477, |
| "eval_loss": 0.05106116086244583, |
| "eval_precision": 0.9905732821395472, |
| "eval_recall": 0.9911244171259008, |
| "eval_runtime": 22.3987, |
| "eval_samples_per_second": 110.363, |
| "eval_steps_per_second": 13.795, |
| "step": 14820 |
| }, |
| { |
| "epoch": 13.0, |
| "learning_rate": 7.446808510638298e-06, |
| "loss": 0.003, |
| "step": 16055 |
| }, |
| { |
| "epoch": 13.0, |
| "eval_accuracy": 0.9939704913450595, |
| "eval_f1": 0.9910981586965161, |
| "eval_loss": 0.050564348697662354, |
| "eval_precision": 0.9910719016584538, |
| "eval_recall": 0.9911244171259008, |
| "eval_runtime": 22.3763, |
| "eval_samples_per_second": 110.474, |
| "eval_steps_per_second": 13.809, |
| "step": 16055 |
| }, |
| { |
| "epoch": 14.0, |
| "learning_rate": 6.382978723404256e-06, |
| "loss": 0.0026, |
| "step": 17290 |
| }, |
| { |
| "epoch": 14.0, |
| "eval_accuracy": 0.9940902829077405, |
| "eval_f1": 0.9915086966312971, |
| "eval_loss": 0.05338473618030548, |
| "eval_precision": 0.9914955623261359, |
| "eval_recall": 0.9915218312844426, |
| "eval_runtime": 22.5218, |
| "eval_samples_per_second": 109.76, |
| "eval_steps_per_second": 13.72, |
| "step": 17290 |
| }, |
| { |
| "epoch": 15.0, |
| "learning_rate": 5.319148936170213e-06, |
| "loss": 0.0019, |
| "step": 18525 |
| }, |
| { |
| "epoch": 15.0, |
| "eval_accuracy": 0.9939105955637191, |
| "eval_f1": 0.9911127004940332, |
| "eval_loss": 0.05444112420082092, |
| "eval_precision": 0.9909420769658607, |
| "eval_recall": 0.9912833827893175, |
| "eval_runtime": 22.3561, |
| "eval_samples_per_second": 110.574, |
| "eval_steps_per_second": 13.822, |
| "step": 18525 |
| }, |
| { |
| "epoch": 16.0, |
| "learning_rate": 4.255319148936171e-06, |
| "loss": 0.0021, |
| "step": 19760 |
| }, |
| { |
| "epoch": 16.0, |
| "eval_accuracy": 0.9932118114480804, |
| "eval_f1": 0.9898781134075251, |
| "eval_loss": 0.05772731453180313, |
| "eval_precision": 0.9899830400678398, |
| "eval_recall": 0.9897732089868588, |
| "eval_runtime": 22.3439, |
| "eval_samples_per_second": 110.634, |
| "eval_steps_per_second": 13.829, |
| "step": 19760 |
| }, |
| { |
| "epoch": 17.0, |
| "learning_rate": 3.191489361702128e-06, |
| "loss": 0.002, |
| "step": 20995 |
| }, |
| { |
| "epoch": 17.0, |
| "eval_accuracy": 0.9941302134286342, |
| "eval_f1": 0.9915220559014439, |
| "eval_loss": 0.054844070225954056, |
| "eval_precision": 0.9914957876331267, |
| "eval_recall": 0.9915483255616787, |
| "eval_runtime": 22.2286, |
| "eval_samples_per_second": 111.208, |
| "eval_steps_per_second": 13.901, |
| "step": 20995 |
| }, |
| { |
| "epoch": 18.0, |
| "learning_rate": 2.1276595744680853e-06, |
| "loss": 0.0019, |
| "step": 22230 |
| }, |
| { |
| "epoch": 18.0, |
| "eval_accuracy": 0.9939704913450595, |
| "eval_f1": 0.9911230059886585, |
| "eval_loss": 0.05662142112851143, |
| "eval_precision": 0.9912806106222835, |
| "eval_recall": 0.990965451462484, |
| "eval_runtime": 22.2877, |
| "eval_samples_per_second": 110.913, |
| "eval_steps_per_second": 13.864, |
| "step": 22230 |
| }, |
| { |
| "epoch": 19.0, |
| "learning_rate": 1.0638297872340427e-06, |
| "loss": 0.0026, |
| "step": 23465 |
| }, |
| { |
| "epoch": 19.0, |
| "eval_accuracy": 0.9933316030107613, |
| "eval_f1": 0.9899420908259677, |
| "eval_loss": 0.05778687819838524, |
| "eval_precision": 0.9902701556244863, |
| "eval_recall": 0.9896142433234422, |
| "eval_runtime": 22.4497, |
| "eval_samples_per_second": 110.113, |
| "eval_steps_per_second": 13.764, |
| "step": 23465 |
| }, |
| { |
| "epoch": 20.0, |
| "learning_rate": 0.0, |
| "loss": 0.0025, |
| "step": 24700 |
| }, |
| { |
| "epoch": 20.0, |
| "eval_accuracy": 0.9933914987921018, |
| "eval_f1": 0.9900625397498409, |
| "eval_loss": 0.057573601603507996, |
| "eval_precision": 0.9902724766751484, |
| "eval_recall": 0.9898526918185672, |
| "eval_runtime": 22.5179, |
| "eval_samples_per_second": 109.779, |
| "eval_steps_per_second": 13.722, |
| "step": 24700 |
| }, |
| { |
| "epoch": 20.0, |
| "step": 24700, |
| "total_flos": 5.441699015795328e+16, |
| "train_loss": 0.10530064259463476, |
| "train_runtime": 11788.9391, |
| "train_samples_per_second": 33.537, |
| "train_steps_per_second": 2.095 |
| } |
| ], |
| "max_steps": 24700, |
| "num_train_epochs": 20, |
| "total_flos": 5.441699015795328e+16, |
| "trial_name": null, |
| "trial_params": null |
| } |
|
|
