| { | |
| "best_metric": 0.9648942420681551, | |
| "best_model_checkpoint": ".//debugged_ru_gsd_ses_udpipe_16_0.01_0.00005_15_04-22-24_21-06/checkpoint-2772", | |
| "epoch": 15.0, | |
| "global_step": 3780, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 4.964818463270476e-05, | |
| "loss": 2.2858, | |
| "step": 252 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_accuracy": 0.8843166071803485, | |
| "eval_f1": 0.8188400388146599, | |
| "eval_loss": 0.5495285391807556, | |
| "eval_precision": 0.8330801944106926, | |
| "eval_recall": 0.8050785263466902, | |
| "eval_runtime": 3.87, | |
| "eval_samples_per_second": 130.234, | |
| "eval_steps_per_second": 16.279, | |
| "step": 252 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "learning_rate": 4.6101885730368706e-05, | |
| "loss": 0.4328, | |
| "step": 504 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_accuracy": 0.9476170480789419, | |
| "eval_f1": 0.9214275223968277, | |
| "eval_loss": 0.25436148047447205, | |
| "eval_precision": 0.9219691403379868, | |
| "eval_recall": 0.920886540437399, | |
| "eval_runtime": 3.854, | |
| "eval_samples_per_second": 130.773, | |
| "eval_steps_per_second": 16.347, | |
| "step": 504 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "learning_rate": 4.255558682803265e-05, | |
| "loss": 0.2312, | |
| "step": 756 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_accuracy": 0.9588494646231367, | |
| "eval_f1": 0.9394761939696281, | |
| "eval_loss": 0.20127034187316895, | |
| "eval_precision": 0.9391317101789382, | |
| "eval_recall": 0.9398209305739028, | |
| "eval_runtime": 3.8428, | |
| "eval_samples_per_second": 131.156, | |
| "eval_steps_per_second": 16.394, | |
| "step": 756 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "learning_rate": 3.9009287925696596e-05, | |
| "loss": 0.1473, | |
| "step": 1008 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "eval_accuracy": 0.9625236195674994, | |
| "eval_f1": 0.9416176470588237, | |
| "eval_loss": 0.1672007292509079, | |
| "eval_precision": 0.9434212465006631, | |
| "eval_recall": 0.9398209305739028, | |
| "eval_runtime": 3.8423, | |
| "eval_samples_per_second": 131.172, | |
| "eval_steps_per_second": 16.396, | |
| "step": 1008 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "learning_rate": 3.5462989023360544e-05, | |
| "loss": 0.1049, | |
| "step": 1260 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "eval_accuracy": 0.9709216880117573, | |
| "eval_f1": 0.9564258946285545, | |
| "eval_loss": 0.15631963312625885, | |
| "eval_precision": 0.9576221306650972, | |
| "eval_recall": 0.9552326434757082, | |
| "eval_runtime": 3.8547, | |
| "eval_samples_per_second": 130.751, | |
| "eval_steps_per_second": 16.344, | |
| "step": 1260 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "learning_rate": 3.191669012102449e-05, | |
| "loss": 0.0836, | |
| "step": 1512 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "eval_accuracy": 0.9724963258450556, | |
| "eval_f1": 0.9572322414427678, | |
| "eval_loss": 0.15163131058216095, | |
| "eval_precision": 0.9601299468399291, | |
| "eval_recall": 0.9543519741670337, | |
| "eval_runtime": 3.8588, | |
| "eval_samples_per_second": 130.609, | |
| "eval_steps_per_second": 16.326, | |
| "step": 1512 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "learning_rate": 2.8370391218688434e-05, | |
| "loss": 0.0506, | |
| "step": 1764 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "eval_accuracy": 0.9718664707117363, | |
| "eval_f1": 0.9565600882028665, | |
| "eval_loss": 0.1509512960910797, | |
| "eval_precision": 0.9580388692579506, | |
| "eval_recall": 0.9550858652575958, | |
| "eval_runtime": 3.8475, | |
| "eval_samples_per_second": 130.995, | |
| "eval_steps_per_second": 16.374, | |
| "step": 1764 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "learning_rate": 2.482409231635238e-05, | |
| "loss": 0.0364, | |
| "step": 2016 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "eval_accuracy": 0.973126180978375, | |
| "eval_f1": 0.9582385321100917, | |
| "eval_loss": 0.15088674426078796, | |
| "eval_precision": 0.9583088667058133, | |
| "eval_recall": 0.9581682078379569, | |
| "eval_runtime": 3.8432, | |
| "eval_samples_per_second": 131.139, | |
| "eval_steps_per_second": 16.392, | |
| "step": 2016 | |
| }, | |
| { | |
| "epoch": 9.0, | |
| "learning_rate": 2.1277793414016324e-05, | |
| "loss": 0.0273, | |
| "step": 2268 | |
| }, | |
| { | |
| "epoch": 9.0, | |
| "eval_accuracy": 0.9728112534117153, | |
| "eval_f1": 0.9574937490807472, | |
| "eval_loss": 0.15090352296829224, | |
| "eval_precision": 0.9594694178334562, | |
| "eval_recall": 0.9555261999119331, | |
| "eval_runtime": 3.8454, | |
| "eval_samples_per_second": 131.066, | |
| "eval_steps_per_second": 16.383, | |
| "step": 2268 | |
| }, | |
| { | |
| "epoch": 10.0, | |
| "learning_rate": 1.7731494511680272e-05, | |
| "loss": 0.0214, | |
| "step": 2520 | |
| }, | |
| { | |
| "epoch": 10.0, | |
| "eval_accuracy": 0.9751207222338862, | |
| "eval_f1": 0.9623209695189129, | |
| "eval_loss": 0.1550149917602539, | |
| "eval_precision": 0.9630990885033813, | |
| "eval_recall": 0.9615441068545428, | |
| "eval_runtime": 3.9336, | |
| "eval_samples_per_second": 128.128, | |
| "eval_steps_per_second": 16.016, | |
| "step": 2520 | |
| }, | |
| { | |
| "epoch": 11.0, | |
| "learning_rate": 1.4185195609344217e-05, | |
| "loss": 0.0138, | |
| "step": 2772 | |
| }, | |
| { | |
| "epoch": 11.0, | |
| "eval_accuracy": 0.976905311778291, | |
| "eval_f1": 0.9648942420681551, | |
| "eval_loss": 0.14702539145946503, | |
| "eval_precision": 0.9656034102601794, | |
| "eval_recall": 0.9641861147805666, | |
| "eval_runtime": 3.9253, | |
| "eval_samples_per_second": 128.398, | |
| "eval_steps_per_second": 16.05, | |
| "step": 2772 | |
| }, | |
| { | |
| "epoch": 12.0, | |
| "learning_rate": 1.0638896707008162e-05, | |
| "loss": 0.0102, | |
| "step": 3024 | |
| }, | |
| { | |
| "epoch": 12.0, | |
| "eval_accuracy": 0.9754356498005459, | |
| "eval_f1": 0.9619005589879376, | |
| "eval_loss": 0.15176355838775635, | |
| "eval_precision": 0.9640277163496977, | |
| "eval_recall": 0.9597827682371936, | |
| "eval_runtime": 3.9451, | |
| "eval_samples_per_second": 127.755, | |
| "eval_steps_per_second": 15.969, | |
| "step": 3024 | |
| }, | |
| { | |
| "epoch": 13.0, | |
| "learning_rate": 7.0925978046721085e-06, | |
| "loss": 0.0084, | |
| "step": 3276 | |
| }, | |
| { | |
| "epoch": 13.0, | |
| "eval_accuracy": 0.9764854083560781, | |
| "eval_f1": 0.9637526652452025, | |
| "eval_loss": 0.15135051310062408, | |
| "eval_precision": 0.9655274012964055, | |
| "eval_recall": 0.9619844415088801, | |
| "eval_runtime": 3.9363, | |
| "eval_samples_per_second": 128.038, | |
| "eval_steps_per_second": 16.005, | |
| "step": 3276 | |
| }, | |
| { | |
| "epoch": 14.0, | |
| "learning_rate": 3.5462989023360543e-06, | |
| "loss": 0.0069, | |
| "step": 3528 | |
| }, | |
| { | |
| "epoch": 14.0, | |
| "eval_accuracy": 0.9765903842116314, | |
| "eval_f1": 0.9642909625275533, | |
| "eval_loss": 0.15399284660816193, | |
| "eval_precision": 0.9654259232014124, | |
| "eval_recall": 0.9631586672537795, | |
| "eval_runtime": 3.9275, | |
| "eval_samples_per_second": 128.326, | |
| "eval_steps_per_second": 16.041, | |
| "step": 3528 | |
| }, | |
| { | |
| "epoch": 15.0, | |
| "learning_rate": 0.0, | |
| "loss": 0.0061, | |
| "step": 3780 | |
| }, | |
| { | |
| "epoch": 15.0, | |
| "eval_accuracy": 0.9766953600671845, | |
| "eval_f1": 0.9647421771705598, | |
| "eval_loss": 0.15449927747249603, | |
| "eval_precision": 0.9655932951036612, | |
| "eval_recall": 0.9638925583443417, | |
| "eval_runtime": 3.8888, | |
| "eval_samples_per_second": 129.601, | |
| "eval_steps_per_second": 16.2, | |
| "step": 3780 | |
| }, | |
| { | |
| "epoch": 15.0, | |
| "step": 3780, | |
| "total_flos": 7859492504967600.0, | |
| "train_loss": 0.23112083008049658, | |
| "train_runtime": 2555.6695, | |
| "train_samples_per_second": 23.624, | |
| "train_steps_per_second": 1.479 | |
| } | |
| ], | |
| "max_steps": 3780, | |
| "num_train_epochs": 15, | |
| "total_flos": 7859492504967600.0, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |