| { | |
| "best_metric": 0.27115458250045776, | |
| "best_model_checkpoint": "./fine_tune_bert_output_LP_FP/checkpoint-852", | |
| "epoch": 14.0, | |
| "global_step": 2982, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 9.999530516431926e-06, | |
| "loss": 0.0227, | |
| "step": 1 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 9.99530516431925e-06, | |
| "loss": 0.0194, | |
| "step": 10 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 9.990610328638498e-06, | |
| "loss": 0.0265, | |
| "step": 20 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 9.985915492957747e-06, | |
| "loss": 0.0265, | |
| "step": 30 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 9.981220657276996e-06, | |
| "loss": 0.0188, | |
| "step": 40 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 9.976525821596245e-06, | |
| "loss": 0.0188, | |
| "step": 50 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 9.971830985915494e-06, | |
| "loss": 0.0313, | |
| "step": 60 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 9.967136150234742e-06, | |
| "loss": 0.0259, | |
| "step": 70 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 9.962441314553991e-06, | |
| "loss": 0.0249, | |
| "step": 80 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 9.95774647887324e-06, | |
| "loss": 0.0333, | |
| "step": 90 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 9.953051643192489e-06, | |
| "loss": 0.0337, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 9.948356807511738e-06, | |
| "loss": 0.0229, | |
| "step": 110 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 9.943661971830986e-06, | |
| "loss": 0.013, | |
| "step": 120 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 9.938967136150237e-06, | |
| "loss": 0.0332, | |
| "step": 130 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 9.934272300469484e-06, | |
| "loss": 0.0221, | |
| "step": 140 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 9.929577464788733e-06, | |
| "loss": 0.0198, | |
| "step": 150 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 9.924882629107983e-06, | |
| "loss": 0.0286, | |
| "step": 160 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 9.92018779342723e-06, | |
| "loss": 0.0451, | |
| "step": 170 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 9.915492957746479e-06, | |
| "loss": 0.0257, | |
| "step": 180 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 9.91079812206573e-06, | |
| "loss": 0.0185, | |
| "step": 190 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 9.906103286384977e-06, | |
| "loss": 0.032, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 9.901408450704227e-06, | |
| "loss": 0.0215, | |
| "step": 210 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_corporation_f1": 0.2831858407079646, | |
| "eval_creative-work_f1": 0.4444444444444445, | |
| "eval_group_f1": 0.2975206611570248, | |
| "eval_location_f1": 0.6853932584269663, | |
| "eval_loss": 0.2913361191749573, | |
| "eval_overall_accuracy": 0.9507192961597908, | |
| "eval_overall_f1": 0.641696750902527, | |
| "eval_overall_precision": 0.7025691699604744, | |
| "eval_overall_recall": 0.590531561461794, | |
| "eval_person_f1": 0.7787878787878788, | |
| "eval_product_f1": 0.4015444015444016, | |
| "eval_runtime": 6.2036, | |
| "eval_samples_per_second": 162.648, | |
| "eval_steps_per_second": 10.317, | |
| "step": 213 | |
| }, | |
| { | |
| "epoch": 1.03, | |
| "learning_rate": 9.896713615023476e-06, | |
| "loss": 0.0237, | |
| "step": 220 | |
| }, | |
| { | |
| "epoch": 1.08, | |
| "learning_rate": 9.892018779342723e-06, | |
| "loss": 0.0166, | |
| "step": 230 | |
| }, | |
| { | |
| "epoch": 1.13, | |
| "learning_rate": 9.887323943661974e-06, | |
| "loss": 0.0144, | |
| "step": 240 | |
| }, | |
| { | |
| "epoch": 1.17, | |
| "learning_rate": 9.882629107981222e-06, | |
| "loss": 0.0198, | |
| "step": 250 | |
| }, | |
| { | |
| "epoch": 1.22, | |
| "learning_rate": 9.87793427230047e-06, | |
| "loss": 0.0299, | |
| "step": 260 | |
| }, | |
| { | |
| "epoch": 1.27, | |
| "learning_rate": 9.87323943661972e-06, | |
| "loss": 0.0123, | |
| "step": 270 | |
| }, | |
| { | |
| "epoch": 1.31, | |
| "learning_rate": 9.868544600938969e-06, | |
| "loss": 0.0143, | |
| "step": 280 | |
| }, | |
| { | |
| "epoch": 1.36, | |
| "learning_rate": 9.863849765258216e-06, | |
| "loss": 0.0227, | |
| "step": 290 | |
| }, | |
| { | |
| "epoch": 1.41, | |
| "learning_rate": 9.859154929577466e-06, | |
| "loss": 0.0125, | |
| "step": 300 | |
| }, | |
| { | |
| "epoch": 1.46, | |
| "learning_rate": 9.854460093896713e-06, | |
| "loss": 0.0374, | |
| "step": 310 | |
| }, | |
| { | |
| "epoch": 1.5, | |
| "learning_rate": 9.849765258215964e-06, | |
| "loss": 0.0126, | |
| "step": 320 | |
| }, | |
| { | |
| "epoch": 1.55, | |
| "learning_rate": 9.845070422535213e-06, | |
| "loss": 0.0234, | |
| "step": 330 | |
| }, | |
| { | |
| "epoch": 1.6, | |
| "learning_rate": 9.84037558685446e-06, | |
| "loss": 0.009, | |
| "step": 340 | |
| }, | |
| { | |
| "epoch": 1.64, | |
| "learning_rate": 9.83568075117371e-06, | |
| "loss": 0.0149, | |
| "step": 350 | |
| }, | |
| { | |
| "epoch": 1.69, | |
| "learning_rate": 9.830985915492959e-06, | |
| "loss": 0.0289, | |
| "step": 360 | |
| }, | |
| { | |
| "epoch": 1.74, | |
| "learning_rate": 9.826291079812206e-06, | |
| "loss": 0.03, | |
| "step": 370 | |
| }, | |
| { | |
| "epoch": 1.78, | |
| "learning_rate": 9.821596244131457e-06, | |
| "loss": 0.0278, | |
| "step": 380 | |
| }, | |
| { | |
| "epoch": 1.83, | |
| "learning_rate": 9.816901408450705e-06, | |
| "loss": 0.0192, | |
| "step": 390 | |
| }, | |
| { | |
| "epoch": 1.88, | |
| "learning_rate": 9.812206572769954e-06, | |
| "loss": 0.0187, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 1.92, | |
| "learning_rate": 9.807511737089203e-06, | |
| "loss": 0.0184, | |
| "step": 410 | |
| }, | |
| { | |
| "epoch": 1.97, | |
| "learning_rate": 9.802816901408452e-06, | |
| "loss": 0.0213, | |
| "step": 420 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_corporation_f1": 0.2830188679245283, | |
| "eval_creative-work_f1": 0.3482587064676617, | |
| "eval_group_f1": 0.3230769230769231, | |
| "eval_location_f1": 0.6857142857142858, | |
| "eval_loss": 0.30522316694259644, | |
| "eval_overall_accuracy": 0.9494709309237903, | |
| "eval_overall_f1": 0.6233183856502242, | |
| "eval_overall_precision": 0.6773879142300195, | |
| "eval_overall_recall": 0.5772425249169435, | |
| "eval_person_f1": 0.7727620504973222, | |
| "eval_product_f1": 0.37942122186495186, | |
| "eval_runtime": 6.2329, | |
| "eval_samples_per_second": 161.883, | |
| "eval_steps_per_second": 10.268, | |
| "step": 426 | |
| }, | |
| { | |
| "epoch": 2.02, | |
| "learning_rate": 9.7981220657277e-06, | |
| "loss": 0.0294, | |
| "step": 430 | |
| }, | |
| { | |
| "epoch": 2.07, | |
| "learning_rate": 9.79342723004695e-06, | |
| "loss": 0.0179, | |
| "step": 440 | |
| }, | |
| { | |
| "epoch": 2.11, | |
| "learning_rate": 9.788732394366198e-06, | |
| "loss": 0.021, | |
| "step": 450 | |
| }, | |
| { | |
| "epoch": 2.16, | |
| "learning_rate": 9.784037558685447e-06, | |
| "loss": 0.0128, | |
| "step": 460 | |
| }, | |
| { | |
| "epoch": 2.21, | |
| "learning_rate": 9.779342723004696e-06, | |
| "loss": 0.0138, | |
| "step": 470 | |
| }, | |
| { | |
| "epoch": 2.25, | |
| "learning_rate": 9.774647887323945e-06, | |
| "loss": 0.0293, | |
| "step": 480 | |
| }, | |
| { | |
| "epoch": 2.3, | |
| "learning_rate": 9.769953051643193e-06, | |
| "loss": 0.0099, | |
| "step": 490 | |
| }, | |
| { | |
| "epoch": 2.35, | |
| "learning_rate": 9.765258215962442e-06, | |
| "loss": 0.0088, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 2.39, | |
| "learning_rate": 9.760563380281691e-06, | |
| "loss": 0.0137, | |
| "step": 510 | |
| }, | |
| { | |
| "epoch": 2.44, | |
| "learning_rate": 9.75586854460094e-06, | |
| "loss": 0.021, | |
| "step": 520 | |
| }, | |
| { | |
| "epoch": 2.49, | |
| "learning_rate": 9.751173708920188e-06, | |
| "loss": 0.0197, | |
| "step": 530 | |
| }, | |
| { | |
| "epoch": 2.54, | |
| "learning_rate": 9.746478873239437e-06, | |
| "loss": 0.0208, | |
| "step": 540 | |
| }, | |
| { | |
| "epoch": 2.58, | |
| "learning_rate": 9.741784037558686e-06, | |
| "loss": 0.0074, | |
| "step": 550 | |
| }, | |
| { | |
| "epoch": 2.63, | |
| "learning_rate": 9.737089201877935e-06, | |
| "loss": 0.017, | |
| "step": 560 | |
| }, | |
| { | |
| "epoch": 2.68, | |
| "learning_rate": 9.732394366197184e-06, | |
| "loss": 0.0205, | |
| "step": 570 | |
| }, | |
| { | |
| "epoch": 2.72, | |
| "learning_rate": 9.727699530516432e-06, | |
| "loss": 0.0159, | |
| "step": 580 | |
| }, | |
| { | |
| "epoch": 2.77, | |
| "learning_rate": 9.723004694835681e-06, | |
| "loss": 0.0139, | |
| "step": 590 | |
| }, | |
| { | |
| "epoch": 2.82, | |
| "learning_rate": 9.71830985915493e-06, | |
| "loss": 0.017, | |
| "step": 600 | |
| }, | |
| { | |
| "epoch": 2.86, | |
| "learning_rate": 9.713615023474179e-06, | |
| "loss": 0.0171, | |
| "step": 610 | |
| }, | |
| { | |
| "epoch": 2.91, | |
| "learning_rate": 9.708920187793428e-06, | |
| "loss": 0.0164, | |
| "step": 620 | |
| }, | |
| { | |
| "epoch": 2.96, | |
| "learning_rate": 9.704225352112678e-06, | |
| "loss": 0.0288, | |
| "step": 630 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_corporation_f1": 0.3076923076923077, | |
| "eval_creative-work_f1": 0.41841004184100417, | |
| "eval_group_f1": 0.3529411764705882, | |
| "eval_location_f1": 0.6222222222222222, | |
| "eval_loss": 0.3378466069698334, | |
| "eval_overall_accuracy": 0.946736416597313, | |
| "eval_overall_f1": 0.6187587494167056, | |
| "eval_overall_precision": 0.7060702875399361, | |
| "eval_overall_recall": 0.5506644518272426, | |
| "eval_person_f1": 0.7532051282051282, | |
| "eval_product_f1": 0.39097744360902253, | |
| "eval_runtime": 2.031, | |
| "eval_samples_per_second": 496.798, | |
| "eval_steps_per_second": 31.511, | |
| "step": 639 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "learning_rate": 9.699530516431925e-06, | |
| "loss": 0.023, | |
| "step": 640 | |
| }, | |
| { | |
| "epoch": 3.05, | |
| "learning_rate": 9.694835680751174e-06, | |
| "loss": 0.0173, | |
| "step": 650 | |
| }, | |
| { | |
| "epoch": 3.1, | |
| "learning_rate": 9.690140845070424e-06, | |
| "loss": 0.0149, | |
| "step": 660 | |
| }, | |
| { | |
| "epoch": 3.15, | |
| "learning_rate": 9.685446009389672e-06, | |
| "loss": 0.0141, | |
| "step": 670 | |
| }, | |
| { | |
| "epoch": 3.19, | |
| "learning_rate": 9.68075117370892e-06, | |
| "loss": 0.0172, | |
| "step": 680 | |
| }, | |
| { | |
| "epoch": 3.24, | |
| "learning_rate": 9.67605633802817e-06, | |
| "loss": 0.0072, | |
| "step": 690 | |
| }, | |
| { | |
| "epoch": 3.29, | |
| "learning_rate": 9.671361502347418e-06, | |
| "loss": 0.015, | |
| "step": 700 | |
| }, | |
| { | |
| "epoch": 3.33, | |
| "learning_rate": 9.666666666666667e-06, | |
| "loss": 0.009, | |
| "step": 710 | |
| }, | |
| { | |
| "epoch": 3.38, | |
| "learning_rate": 9.661971830985917e-06, | |
| "loss": 0.0098, | |
| "step": 720 | |
| }, | |
| { | |
| "epoch": 3.43, | |
| "learning_rate": 9.657276995305164e-06, | |
| "loss": 0.018, | |
| "step": 730 | |
| }, | |
| { | |
| "epoch": 3.47, | |
| "learning_rate": 9.652582159624415e-06, | |
| "loss": 0.0092, | |
| "step": 740 | |
| }, | |
| { | |
| "epoch": 3.52, | |
| "learning_rate": 9.647887323943664e-06, | |
| "loss": 0.0166, | |
| "step": 750 | |
| }, | |
| { | |
| "epoch": 3.57, | |
| "learning_rate": 9.64319248826291e-06, | |
| "loss": 0.0085, | |
| "step": 760 | |
| }, | |
| { | |
| "epoch": 3.62, | |
| "learning_rate": 9.638497652582161e-06, | |
| "loss": 0.0168, | |
| "step": 770 | |
| }, | |
| { | |
| "epoch": 3.66, | |
| "learning_rate": 9.63380281690141e-06, | |
| "loss": 0.0201, | |
| "step": 780 | |
| }, | |
| { | |
| "epoch": 3.71, | |
| "learning_rate": 9.629107981220657e-06, | |
| "loss": 0.0143, | |
| "step": 790 | |
| }, | |
| { | |
| "epoch": 3.76, | |
| "learning_rate": 9.624413145539908e-06, | |
| "loss": 0.0189, | |
| "step": 800 | |
| }, | |
| { | |
| "epoch": 3.8, | |
| "learning_rate": 9.619718309859156e-06, | |
| "loss": 0.0099, | |
| "step": 810 | |
| }, | |
| { | |
| "epoch": 3.85, | |
| "learning_rate": 9.615023474178405e-06, | |
| "loss": 0.0259, | |
| "step": 820 | |
| }, | |
| { | |
| "epoch": 3.9, | |
| "learning_rate": 9.610328638497654e-06, | |
| "loss": 0.0187, | |
| "step": 830 | |
| }, | |
| { | |
| "epoch": 3.94, | |
| "learning_rate": 9.605633802816903e-06, | |
| "loss": 0.0102, | |
| "step": 840 | |
| }, | |
| { | |
| "epoch": 3.99, | |
| "learning_rate": 9.600938967136152e-06, | |
| "loss": 0.0124, | |
| "step": 850 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "eval_corporation_f1": 0.30769230769230765, | |
| "eval_creative-work_f1": 0.48421052631578954, | |
| "eval_group_f1": 0.31666666666666665, | |
| "eval_location_f1": 0.6808510638297872, | |
| "eval_loss": 0.27115458250045776, | |
| "eval_overall_accuracy": 0.9502437284508382, | |
| "eval_overall_f1": 0.6339784946236559, | |
| "eval_overall_precision": 0.6574487065120428, | |
| "eval_overall_recall": 0.6121262458471761, | |
| "eval_person_f1": 0.7734553775743707, | |
| "eval_product_f1": 0.3986254295532646, | |
| "eval_runtime": 5.0216, | |
| "eval_samples_per_second": 200.934, | |
| "eval_steps_per_second": 12.745, | |
| "step": 852 | |
| }, | |
| { | |
| "epoch": 4.04, | |
| "learning_rate": 9.5962441314554e-06, | |
| "loss": 0.017, | |
| "step": 860 | |
| }, | |
| { | |
| "epoch": 4.08, | |
| "learning_rate": 9.591549295774649e-06, | |
| "loss": 0.0138, | |
| "step": 870 | |
| }, | |
| { | |
| "epoch": 4.13, | |
| "learning_rate": 9.586854460093898e-06, | |
| "loss": 0.0122, | |
| "step": 880 | |
| }, | |
| { | |
| "epoch": 4.18, | |
| "learning_rate": 9.582159624413147e-06, | |
| "loss": 0.0116, | |
| "step": 890 | |
| }, | |
| { | |
| "epoch": 4.23, | |
| "learning_rate": 9.577464788732394e-06, | |
| "loss": 0.0066, | |
| "step": 900 | |
| }, | |
| { | |
| "epoch": 4.27, | |
| "learning_rate": 9.572769953051644e-06, | |
| "loss": 0.016, | |
| "step": 910 | |
| }, | |
| { | |
| "epoch": 4.32, | |
| "learning_rate": 9.568075117370893e-06, | |
| "loss": 0.0203, | |
| "step": 920 | |
| }, | |
| { | |
| "epoch": 4.37, | |
| "learning_rate": 9.563380281690142e-06, | |
| "loss": 0.0092, | |
| "step": 930 | |
| }, | |
| { | |
| "epoch": 4.41, | |
| "learning_rate": 9.55868544600939e-06, | |
| "loss": 0.0098, | |
| "step": 940 | |
| }, | |
| { | |
| "epoch": 4.46, | |
| "learning_rate": 9.55399061032864e-06, | |
| "loss": 0.0083, | |
| "step": 950 | |
| }, | |
| { | |
| "epoch": 4.51, | |
| "learning_rate": 9.549295774647888e-06, | |
| "loss": 0.0244, | |
| "step": 960 | |
| }, | |
| { | |
| "epoch": 4.55, | |
| "learning_rate": 9.544600938967137e-06, | |
| "loss": 0.0096, | |
| "step": 970 | |
| }, | |
| { | |
| "epoch": 4.6, | |
| "learning_rate": 9.539906103286386e-06, | |
| "loss": 0.0109, | |
| "step": 980 | |
| }, | |
| { | |
| "epoch": 4.65, | |
| "learning_rate": 9.535211267605635e-06, | |
| "loss": 0.0114, | |
| "step": 990 | |
| }, | |
| { | |
| "epoch": 4.69, | |
| "learning_rate": 9.530516431924883e-06, | |
| "loss": 0.0214, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 4.74, | |
| "learning_rate": 9.525821596244132e-06, | |
| "loss": 0.0061, | |
| "step": 1010 | |
| }, | |
| { | |
| "epoch": 4.79, | |
| "learning_rate": 9.521126760563381e-06, | |
| "loss": 0.011, | |
| "step": 1020 | |
| }, | |
| { | |
| "epoch": 4.84, | |
| "learning_rate": 9.51643192488263e-06, | |
| "loss": 0.0061, | |
| "step": 1030 | |
| }, | |
| { | |
| "epoch": 4.88, | |
| "learning_rate": 9.511737089201879e-06, | |
| "loss": 0.0202, | |
| "step": 1040 | |
| }, | |
| { | |
| "epoch": 4.93, | |
| "learning_rate": 9.507042253521127e-06, | |
| "loss": 0.0249, | |
| "step": 1050 | |
| }, | |
| { | |
| "epoch": 4.98, | |
| "learning_rate": 9.502347417840376e-06, | |
| "loss": 0.0208, | |
| "step": 1060 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "eval_corporation_f1": 0.3063063063063063, | |
| "eval_creative-work_f1": 0.42857142857142855, | |
| "eval_group_f1": 0.3418803418803419, | |
| "eval_location_f1": 0.7052023121387283, | |
| "eval_loss": 0.29045674204826355, | |
| "eval_overall_accuracy": 0.9518487694685531, | |
| "eval_overall_f1": 0.6544150605109816, | |
| "eval_overall_precision": 0.7108081791626095, | |
| "eval_overall_recall": 0.606312292358804, | |
| "eval_person_f1": 0.7912584777694045, | |
| "eval_product_f1": 0.42231075697211157, | |
| "eval_runtime": 4.8738, | |
| "eval_samples_per_second": 207.026, | |
| "eval_steps_per_second": 13.131, | |
| "step": 1065 | |
| }, | |
| { | |
| "epoch": 5.02, | |
| "learning_rate": 9.497652582159625e-06, | |
| "loss": 0.0067, | |
| "step": 1070 | |
| }, | |
| { | |
| "epoch": 5.07, | |
| "learning_rate": 9.492957746478874e-06, | |
| "loss": 0.017, | |
| "step": 1080 | |
| }, | |
| { | |
| "epoch": 5.12, | |
| "learning_rate": 9.488262910798123e-06, | |
| "loss": 0.0069, | |
| "step": 1090 | |
| }, | |
| { | |
| "epoch": 5.16, | |
| "learning_rate": 9.483568075117371e-06, | |
| "loss": 0.0109, | |
| "step": 1100 | |
| }, | |
| { | |
| "epoch": 5.21, | |
| "learning_rate": 9.47887323943662e-06, | |
| "loss": 0.0108, | |
| "step": 1110 | |
| }, | |
| { | |
| "epoch": 5.26, | |
| "learning_rate": 9.474178403755869e-06, | |
| "loss": 0.0112, | |
| "step": 1120 | |
| }, | |
| { | |
| "epoch": 5.31, | |
| "learning_rate": 9.469483568075118e-06, | |
| "loss": 0.0042, | |
| "step": 1130 | |
| }, | |
| { | |
| "epoch": 5.35, | |
| "learning_rate": 9.464788732394366e-06, | |
| "loss": 0.0086, | |
| "step": 1140 | |
| }, | |
| { | |
| "epoch": 5.4, | |
| "learning_rate": 9.460093896713615e-06, | |
| "loss": 0.0067, | |
| "step": 1150 | |
| }, | |
| { | |
| "epoch": 5.45, | |
| "learning_rate": 9.455399061032866e-06, | |
| "loss": 0.0099, | |
| "step": 1160 | |
| }, | |
| { | |
| "epoch": 5.49, | |
| "learning_rate": 9.450704225352113e-06, | |
| "loss": 0.0071, | |
| "step": 1170 | |
| }, | |
| { | |
| "epoch": 5.54, | |
| "learning_rate": 9.446009389671362e-06, | |
| "loss": 0.0061, | |
| "step": 1180 | |
| }, | |
| { | |
| "epoch": 5.59, | |
| "learning_rate": 9.441314553990612e-06, | |
| "loss": 0.0088, | |
| "step": 1190 | |
| }, | |
| { | |
| "epoch": 5.63, | |
| "learning_rate": 9.43661971830986e-06, | |
| "loss": 0.0078, | |
| "step": 1200 | |
| }, | |
| { | |
| "epoch": 5.68, | |
| "learning_rate": 9.431924882629108e-06, | |
| "loss": 0.0041, | |
| "step": 1210 | |
| }, | |
| { | |
| "epoch": 5.73, | |
| "learning_rate": 9.427230046948358e-06, | |
| "loss": 0.019, | |
| "step": 1220 | |
| }, | |
| { | |
| "epoch": 5.77, | |
| "learning_rate": 9.422535211267606e-06, | |
| "loss": 0.0113, | |
| "step": 1230 | |
| }, | |
| { | |
| "epoch": 5.82, | |
| "learning_rate": 9.417840375586856e-06, | |
| "loss": 0.0245, | |
| "step": 1240 | |
| }, | |
| { | |
| "epoch": 5.87, | |
| "learning_rate": 9.413145539906105e-06, | |
| "loss": 0.0195, | |
| "step": 1250 | |
| }, | |
| { | |
| "epoch": 5.92, | |
| "learning_rate": 9.408450704225352e-06, | |
| "loss": 0.007, | |
| "step": 1260 | |
| }, | |
| { | |
| "epoch": 5.96, | |
| "learning_rate": 9.403755868544602e-06, | |
| "loss": 0.0071, | |
| "step": 1270 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "eval_corporation_f1": 0.27586206896551724, | |
| "eval_creative-work_f1": 0.4380165289256198, | |
| "eval_group_f1": 0.32558139534883723, | |
| "eval_location_f1": 0.6744186046511628, | |
| "eval_loss": 0.3188508152961731, | |
| "eval_overall_accuracy": 0.9494114849601712, | |
| "eval_overall_f1": 0.6268922528940339, | |
| "eval_overall_precision": 0.6756238003838771, | |
| "eval_overall_recall": 0.584717607973422, | |
| "eval_person_f1": 0.7781250000000001, | |
| "eval_product_f1": 0.37785016286644946, | |
| "eval_runtime": 5.2125, | |
| "eval_samples_per_second": 193.573, | |
| "eval_steps_per_second": 12.278, | |
| "step": 1278 | |
| }, | |
| { | |
| "epoch": 6.01, | |
| "learning_rate": 9.399061032863851e-06, | |
| "loss": 0.0069, | |
| "step": 1280 | |
| }, | |
| { | |
| "epoch": 6.06, | |
| "learning_rate": 9.394366197183098e-06, | |
| "loss": 0.0062, | |
| "step": 1290 | |
| }, | |
| { | |
| "epoch": 6.1, | |
| "learning_rate": 9.389671361502349e-06, | |
| "loss": 0.0096, | |
| "step": 1300 | |
| }, | |
| { | |
| "epoch": 6.15, | |
| "learning_rate": 9.384976525821598e-06, | |
| "loss": 0.0184, | |
| "step": 1310 | |
| }, | |
| { | |
| "epoch": 6.2, | |
| "learning_rate": 9.380281690140845e-06, | |
| "loss": 0.0116, | |
| "step": 1320 | |
| }, | |
| { | |
| "epoch": 6.24, | |
| "learning_rate": 9.375586854460095e-06, | |
| "loss": 0.0092, | |
| "step": 1330 | |
| }, | |
| { | |
| "epoch": 6.29, | |
| "learning_rate": 9.370892018779344e-06, | |
| "loss": 0.0075, | |
| "step": 1340 | |
| }, | |
| { | |
| "epoch": 6.34, | |
| "learning_rate": 9.366197183098593e-06, | |
| "loss": 0.0101, | |
| "step": 1350 | |
| }, | |
| { | |
| "epoch": 6.38, | |
| "learning_rate": 9.361502347417842e-06, | |
| "loss": 0.0106, | |
| "step": 1360 | |
| }, | |
| { | |
| "epoch": 6.43, | |
| "learning_rate": 9.35680751173709e-06, | |
| "loss": 0.0069, | |
| "step": 1370 | |
| }, | |
| { | |
| "epoch": 6.48, | |
| "learning_rate": 9.35211267605634e-06, | |
| "loss": 0.0064, | |
| "step": 1380 | |
| }, | |
| { | |
| "epoch": 6.53, | |
| "learning_rate": 9.347417840375588e-06, | |
| "loss": 0.0112, | |
| "step": 1390 | |
| }, | |
| { | |
| "epoch": 6.57, | |
| "learning_rate": 9.342723004694837e-06, | |
| "loss": 0.0059, | |
| "step": 1400 | |
| }, | |
| { | |
| "epoch": 6.62, | |
| "learning_rate": 9.338028169014086e-06, | |
| "loss": 0.006, | |
| "step": 1410 | |
| }, | |
| { | |
| "epoch": 6.67, | |
| "learning_rate": 9.333333333333334e-06, | |
| "loss": 0.0089, | |
| "step": 1420 | |
| }, | |
| { | |
| "epoch": 6.71, | |
| "learning_rate": 9.328638497652583e-06, | |
| "loss": 0.0068, | |
| "step": 1430 | |
| }, | |
| { | |
| "epoch": 6.76, | |
| "learning_rate": 9.323943661971832e-06, | |
| "loss": 0.0041, | |
| "step": 1440 | |
| }, | |
| { | |
| "epoch": 6.81, | |
| "learning_rate": 9.31924882629108e-06, | |
| "loss": 0.0191, | |
| "step": 1450 | |
| }, | |
| { | |
| "epoch": 6.85, | |
| "learning_rate": 9.31455399061033e-06, | |
| "loss": 0.0123, | |
| "step": 1460 | |
| }, | |
| { | |
| "epoch": 6.9, | |
| "learning_rate": 9.309859154929578e-06, | |
| "loss": 0.0033, | |
| "step": 1470 | |
| }, | |
| { | |
| "epoch": 6.95, | |
| "learning_rate": 9.305164319248827e-06, | |
| "loss": 0.0085, | |
| "step": 1480 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "learning_rate": 9.300469483568076e-06, | |
| "loss": 0.0073, | |
| "step": 1490 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "eval_corporation_f1": 0.3061224489795918, | |
| "eval_creative-work_f1": 0.4388185654008439, | |
| "eval_group_f1": 0.37837837837837834, | |
| "eval_location_f1": 0.6946107784431138, | |
| "eval_loss": 0.35933107137680054, | |
| "eval_overall_accuracy": 0.94756866008798, | |
| "eval_overall_f1": 0.631031220435194, | |
| "eval_overall_precision": 0.7329670329670329, | |
| "eval_overall_recall": 0.5539867109634552, | |
| "eval_person_f1": 0.7631160572337042, | |
| "eval_product_f1": 0.3374485596707819, | |
| "eval_runtime": 4.8206, | |
| "eval_samples_per_second": 209.311, | |
| "eval_steps_per_second": 13.276, | |
| "step": 1491 | |
| }, | |
| { | |
| "epoch": 7.04, | |
| "learning_rate": 9.295774647887325e-06, | |
| "loss": 0.0109, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 7.09, | |
| "learning_rate": 9.291079812206573e-06, | |
| "loss": 0.0049, | |
| "step": 1510 | |
| }, | |
| { | |
| "epoch": 7.14, | |
| "learning_rate": 9.286384976525822e-06, | |
| "loss": 0.0098, | |
| "step": 1520 | |
| }, | |
| { | |
| "epoch": 7.18, | |
| "learning_rate": 9.281690140845071e-06, | |
| "loss": 0.0053, | |
| "step": 1530 | |
| }, | |
| { | |
| "epoch": 7.23, | |
| "learning_rate": 9.27699530516432e-06, | |
| "loss": 0.0054, | |
| "step": 1540 | |
| }, | |
| { | |
| "epoch": 7.28, | |
| "learning_rate": 9.272300469483569e-06, | |
| "loss": 0.0123, | |
| "step": 1550 | |
| }, | |
| { | |
| "epoch": 7.32, | |
| "learning_rate": 9.267605633802817e-06, | |
| "loss": 0.0037, | |
| "step": 1560 | |
| }, | |
| { | |
| "epoch": 7.37, | |
| "learning_rate": 9.262910798122066e-06, | |
| "loss": 0.0029, | |
| "step": 1570 | |
| }, | |
| { | |
| "epoch": 7.42, | |
| "learning_rate": 9.258215962441315e-06, | |
| "loss": 0.0186, | |
| "step": 1580 | |
| }, | |
| { | |
| "epoch": 7.46, | |
| "learning_rate": 9.253521126760564e-06, | |
| "loss": 0.0089, | |
| "step": 1590 | |
| }, | |
| { | |
| "epoch": 7.51, | |
| "learning_rate": 9.248826291079813e-06, | |
| "loss": 0.006, | |
| "step": 1600 | |
| }, | |
| { | |
| "epoch": 7.56, | |
| "learning_rate": 9.244131455399061e-06, | |
| "loss": 0.0087, | |
| "step": 1610 | |
| }, | |
| { | |
| "epoch": 7.61, | |
| "learning_rate": 9.23943661971831e-06, | |
| "loss": 0.0099, | |
| "step": 1620 | |
| }, | |
| { | |
| "epoch": 7.65, | |
| "learning_rate": 9.234741784037559e-06, | |
| "loss": 0.0293, | |
| "step": 1630 | |
| }, | |
| { | |
| "epoch": 7.7, | |
| "learning_rate": 9.230046948356808e-06, | |
| "loss": 0.0129, | |
| "step": 1640 | |
| }, | |
| { | |
| "epoch": 7.75, | |
| "learning_rate": 9.225352112676057e-06, | |
| "loss": 0.0207, | |
| "step": 1650 | |
| }, | |
| { | |
| "epoch": 7.79, | |
| "learning_rate": 9.220657276995307e-06, | |
| "loss": 0.0055, | |
| "step": 1660 | |
| }, | |
| { | |
| "epoch": 7.84, | |
| "learning_rate": 9.215962441314554e-06, | |
| "loss": 0.0059, | |
| "step": 1670 | |
| }, | |
| { | |
| "epoch": 7.89, | |
| "learning_rate": 9.211267605633803e-06, | |
| "loss": 0.0103, | |
| "step": 1680 | |
| }, | |
| { | |
| "epoch": 7.93, | |
| "learning_rate": 9.206572769953053e-06, | |
| "loss": 0.018, | |
| "step": 1690 | |
| }, | |
| { | |
| "epoch": 7.98, | |
| "learning_rate": 9.2018779342723e-06, | |
| "loss": 0.0135, | |
| "step": 1700 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "eval_corporation_f1": 0.34, | |
| "eval_creative-work_f1": 0.417910447761194, | |
| "eval_group_f1": 0.3088235294117647, | |
| "eval_location_f1": 0.6631578947368422, | |
| "eval_loss": 0.35640954971313477, | |
| "eval_overall_accuracy": 0.9470930923790275, | |
| "eval_overall_f1": 0.609981515711645, | |
| "eval_overall_precision": 0.6875, | |
| "eval_overall_recall": 0.5481727574750831, | |
| "eval_person_f1": 0.7485667485667485, | |
| "eval_product_f1": 0.3694779116465864, | |
| "eval_runtime": 6.3558, | |
| "eval_samples_per_second": 158.753, | |
| "eval_steps_per_second": 10.07, | |
| "step": 1704 | |
| }, | |
| { | |
| "epoch": 8.03, | |
| "learning_rate": 9.19718309859155e-06, | |
| "loss": 0.006, | |
| "step": 1710 | |
| }, | |
| { | |
| "epoch": 8.08, | |
| "learning_rate": 9.1924882629108e-06, | |
| "loss": 0.0065, | |
| "step": 1720 | |
| }, | |
| { | |
| "epoch": 8.12, | |
| "learning_rate": 9.187793427230047e-06, | |
| "loss": 0.014, | |
| "step": 1730 | |
| }, | |
| { | |
| "epoch": 8.17, | |
| "learning_rate": 9.183098591549296e-06, | |
| "loss": 0.0077, | |
| "step": 1740 | |
| }, | |
| { | |
| "epoch": 8.22, | |
| "learning_rate": 9.178403755868546e-06, | |
| "loss": 0.006, | |
| "step": 1750 | |
| }, | |
| { | |
| "epoch": 8.26, | |
| "learning_rate": 9.173708920187793e-06, | |
| "loss": 0.0062, | |
| "step": 1760 | |
| }, | |
| { | |
| "epoch": 8.31, | |
| "learning_rate": 9.169014084507044e-06, | |
| "loss": 0.0082, | |
| "step": 1770 | |
| }, | |
| { | |
| "epoch": 8.36, | |
| "learning_rate": 9.164319248826293e-06, | |
| "loss": 0.0064, | |
| "step": 1780 | |
| }, | |
| { | |
| "epoch": 8.4, | |
| "learning_rate": 9.15962441314554e-06, | |
| "loss": 0.0089, | |
| "step": 1790 | |
| }, | |
| { | |
| "epoch": 8.45, | |
| "learning_rate": 9.15492957746479e-06, | |
| "loss": 0.004, | |
| "step": 1800 | |
| }, | |
| { | |
| "epoch": 8.5, | |
| "learning_rate": 9.150234741784039e-06, | |
| "loss": 0.0082, | |
| "step": 1810 | |
| }, | |
| { | |
| "epoch": 8.54, | |
| "learning_rate": 9.145539906103286e-06, | |
| "loss": 0.0129, | |
| "step": 1820 | |
| }, | |
| { | |
| "epoch": 8.59, | |
| "learning_rate": 9.140845070422536e-06, | |
| "loss": 0.0141, | |
| "step": 1830 | |
| }, | |
| { | |
| "epoch": 8.64, | |
| "learning_rate": 9.136150234741785e-06, | |
| "loss": 0.0121, | |
| "step": 1840 | |
| }, | |
| { | |
| "epoch": 8.69, | |
| "learning_rate": 9.131455399061034e-06, | |
| "loss": 0.0054, | |
| "step": 1850 | |
| }, | |
| { | |
| "epoch": 8.73, | |
| "learning_rate": 9.126760563380283e-06, | |
| "loss": 0.0136, | |
| "step": 1860 | |
| }, | |
| { | |
| "epoch": 8.78, | |
| "learning_rate": 9.122065727699532e-06, | |
| "loss": 0.0036, | |
| "step": 1870 | |
| }, | |
| { | |
| "epoch": 8.83, | |
| "learning_rate": 9.11737089201878e-06, | |
| "loss": 0.0065, | |
| "step": 1880 | |
| }, | |
| { | |
| "epoch": 8.87, | |
| "learning_rate": 9.11267605633803e-06, | |
| "loss": 0.0079, | |
| "step": 1890 | |
| }, | |
| { | |
| "epoch": 8.92, | |
| "learning_rate": 9.107981220657278e-06, | |
| "loss": 0.0154, | |
| "step": 1900 | |
| }, | |
| { | |
| "epoch": 8.97, | |
| "learning_rate": 9.103286384976527e-06, | |
| "loss": 0.0097, | |
| "step": 1910 | |
| }, | |
| { | |
| "epoch": 9.0, | |
| "eval_corporation_f1": 0.3111111111111111, | |
| "eval_creative-work_f1": 0.46093750000000006, | |
| "eval_group_f1": 0.3835616438356164, | |
| "eval_location_f1": 0.7089947089947088, | |
| "eval_loss": 0.30845504999160767, | |
| "eval_overall_accuracy": 0.9515515396504577, | |
| "eval_overall_f1": 0.6495149725854069, | |
| "eval_overall_precision": 0.6598114824335904, | |
| "eval_overall_recall": 0.6395348837209303, | |
| "eval_person_f1": 0.7905604719764011, | |
| "eval_product_f1": 0.40830449826989623, | |
| "eval_runtime": 5.2314, | |
| "eval_samples_per_second": 192.875, | |
| "eval_steps_per_second": 12.234, | |
| "step": 1917 | |
| }, | |
| { | |
| "epoch": 9.01, | |
| "learning_rate": 9.098591549295776e-06, | |
| "loss": 0.005, | |
| "step": 1920 | |
| }, | |
| { | |
| "epoch": 9.06, | |
| "learning_rate": 9.093896713615024e-06, | |
| "loss": 0.0036, | |
| "step": 1930 | |
| }, | |
| { | |
| "epoch": 9.11, | |
| "learning_rate": 9.089201877934273e-06, | |
| "loss": 0.0157, | |
| "step": 1940 | |
| }, | |
| { | |
| "epoch": 9.15, | |
| "learning_rate": 9.084507042253522e-06, | |
| "loss": 0.0044, | |
| "step": 1950 | |
| }, | |
| { | |
| "epoch": 9.2, | |
| "learning_rate": 9.07981220657277e-06, | |
| "loss": 0.0139, | |
| "step": 1960 | |
| }, | |
| { | |
| "epoch": 9.25, | |
| "learning_rate": 9.07511737089202e-06, | |
| "loss": 0.0135, | |
| "step": 1970 | |
| }, | |
| { | |
| "epoch": 9.3, | |
| "learning_rate": 9.070422535211268e-06, | |
| "loss": 0.003, | |
| "step": 1980 | |
| }, | |
| { | |
| "epoch": 9.34, | |
| "learning_rate": 9.065727699530517e-06, | |
| "loss": 0.0078, | |
| "step": 1990 | |
| }, | |
| { | |
| "epoch": 9.39, | |
| "learning_rate": 9.061032863849766e-06, | |
| "loss": 0.0052, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 9.44, | |
| "learning_rate": 9.056338028169015e-06, | |
| "loss": 0.0043, | |
| "step": 2010 | |
| }, | |
| { | |
| "epoch": 9.48, | |
| "learning_rate": 9.051643192488264e-06, | |
| "loss": 0.0075, | |
| "step": 2020 | |
| }, | |
| { | |
| "epoch": 9.53, | |
| "learning_rate": 9.046948356807512e-06, | |
| "loss": 0.0041, | |
| "step": 2030 | |
| }, | |
| { | |
| "epoch": 9.58, | |
| "learning_rate": 9.042253521126761e-06, | |
| "loss": 0.0042, | |
| "step": 2040 | |
| }, | |
| { | |
| "epoch": 9.62, | |
| "learning_rate": 9.03755868544601e-06, | |
| "loss": 0.0064, | |
| "step": 2050 | |
| }, | |
| { | |
| "epoch": 9.67, | |
| "learning_rate": 9.032863849765259e-06, | |
| "loss": 0.0045, | |
| "step": 2060 | |
| }, | |
| { | |
| "epoch": 9.72, | |
| "learning_rate": 9.028169014084507e-06, | |
| "loss": 0.0074, | |
| "step": 2070 | |
| }, | |
| { | |
| "epoch": 9.77, | |
| "learning_rate": 9.023474178403756e-06, | |
| "loss": 0.0048, | |
| "step": 2080 | |
| }, | |
| { | |
| "epoch": 9.81, | |
| "learning_rate": 9.018779342723005e-06, | |
| "loss": 0.0027, | |
| "step": 2090 | |
| }, | |
| { | |
| "epoch": 9.86, | |
| "learning_rate": 9.014084507042254e-06, | |
| "loss": 0.0028, | |
| "step": 2100 | |
| }, | |
| { | |
| "epoch": 9.91, | |
| "learning_rate": 9.009389671361503e-06, | |
| "loss": 0.0096, | |
| "step": 2110 | |
| }, | |
| { | |
| "epoch": 9.95, | |
| "learning_rate": 9.004694835680751e-06, | |
| "loss": 0.0058, | |
| "step": 2120 | |
| }, | |
| { | |
| "epoch": 10.0, | |
| "learning_rate": 9e-06, | |
| "loss": 0.0108, | |
| "step": 2130 | |
| }, | |
| { | |
| "epoch": 10.0, | |
| "eval_corporation_f1": 0.35294117647058826, | |
| "eval_creative-work_f1": 0.45801526717557256, | |
| "eval_group_f1": 0.36486486486486486, | |
| "eval_location_f1": 0.689655172413793, | |
| "eval_loss": 0.30454060435295105, | |
| "eval_overall_accuracy": 0.950897634050648, | |
| "eval_overall_f1": 0.6540880503144654, | |
| "eval_overall_precision": 0.6604572396274344, | |
| "eval_overall_recall": 0.6478405315614618, | |
| "eval_person_f1": 0.784313725490196, | |
| "eval_product_f1": 0.43870967741935485, | |
| "eval_runtime": 6.5888, | |
| "eval_samples_per_second": 153.138, | |
| "eval_steps_per_second": 9.713, | |
| "step": 2130 | |
| }, | |
| { | |
| "epoch": 10.05, | |
| "learning_rate": 8.995305164319249e-06, | |
| "loss": 0.0035, | |
| "step": 2140 | |
| }, | |
| { | |
| "epoch": 10.09, | |
| "learning_rate": 8.990610328638498e-06, | |
| "loss": 0.003, | |
| "step": 2150 | |
| }, | |
| { | |
| "epoch": 10.14, | |
| "learning_rate": 8.985915492957748e-06, | |
| "loss": 0.0117, | |
| "step": 2160 | |
| }, | |
| { | |
| "epoch": 10.19, | |
| "learning_rate": 8.981220657276995e-06, | |
| "loss": 0.004, | |
| "step": 2170 | |
| }, | |
| { | |
| "epoch": 10.23, | |
| "learning_rate": 8.976525821596244e-06, | |
| "loss": 0.0038, | |
| "step": 2180 | |
| }, | |
| { | |
| "epoch": 10.28, | |
| "learning_rate": 8.971830985915495e-06, | |
| "loss": 0.0057, | |
| "step": 2190 | |
| }, | |
| { | |
| "epoch": 10.33, | |
| "learning_rate": 8.967136150234742e-06, | |
| "loss": 0.0025, | |
| "step": 2200 | |
| }, | |
| { | |
| "epoch": 10.38, | |
| "learning_rate": 8.96244131455399e-06, | |
| "loss": 0.0035, | |
| "step": 2210 | |
| }, | |
| { | |
| "epoch": 10.42, | |
| "learning_rate": 8.957746478873241e-06, | |
| "loss": 0.008, | |
| "step": 2220 | |
| }, | |
| { | |
| "epoch": 10.47, | |
| "learning_rate": 8.953051643192488e-06, | |
| "loss": 0.0025, | |
| "step": 2230 | |
| }, | |
| { | |
| "epoch": 10.52, | |
| "learning_rate": 8.948356807511737e-06, | |
| "loss": 0.0023, | |
| "step": 2240 | |
| }, | |
| { | |
| "epoch": 10.56, | |
| "learning_rate": 8.943661971830987e-06, | |
| "loss": 0.0086, | |
| "step": 2250 | |
| }, | |
| { | |
| "epoch": 10.61, | |
| "learning_rate": 8.938967136150235e-06, | |
| "loss": 0.0101, | |
| "step": 2260 | |
| }, | |
| { | |
| "epoch": 10.66, | |
| "learning_rate": 8.934272300469485e-06, | |
| "loss": 0.0072, | |
| "step": 2270 | |
| }, | |
| { | |
| "epoch": 10.7, | |
| "learning_rate": 8.929577464788734e-06, | |
| "loss": 0.004, | |
| "step": 2280 | |
| }, | |
| { | |
| "epoch": 10.75, | |
| "learning_rate": 8.924882629107981e-06, | |
| "loss": 0.0026, | |
| "step": 2290 | |
| }, | |
| { | |
| "epoch": 10.8, | |
| "learning_rate": 8.920187793427231e-06, | |
| "loss": 0.0047, | |
| "step": 2300 | |
| }, | |
| { | |
| "epoch": 10.85, | |
| "learning_rate": 8.91549295774648e-06, | |
| "loss": 0.002, | |
| "step": 2310 | |
| }, | |
| { | |
| "epoch": 10.89, | |
| "learning_rate": 8.910798122065727e-06, | |
| "loss": 0.0116, | |
| "step": 2320 | |
| }, | |
| { | |
| "epoch": 10.94, | |
| "learning_rate": 8.906103286384978e-06, | |
| "loss": 0.0057, | |
| "step": 2330 | |
| }, | |
| { | |
| "epoch": 10.99, | |
| "learning_rate": 8.901408450704227e-06, | |
| "loss": 0.013, | |
| "step": 2340 | |
| }, | |
| { | |
| "epoch": 11.0, | |
| "eval_corporation_f1": 0.2782608695652174, | |
| "eval_creative-work_f1": 0.4247787610619469, | |
| "eval_group_f1": 0.3357664233576642, | |
| "eval_location_f1": 0.736842105263158, | |
| "eval_loss": 0.3382622003555298, | |
| "eval_overall_accuracy": 0.9507192961597908, | |
| "eval_overall_f1": 0.6469565217391304, | |
| "eval_overall_precision": 0.6788321167883211, | |
| "eval_overall_recall": 0.6179401993355482, | |
| "eval_person_f1": 0.7958271236959762, | |
| "eval_product_f1": 0.3655172413793103, | |
| "eval_runtime": 5.3383, | |
| "eval_samples_per_second": 189.013, | |
| "eval_steps_per_second": 11.989, | |
| "step": 2343 | |
| }, | |
| { | |
| "epoch": 11.03, | |
| "learning_rate": 8.896713615023475e-06, | |
| "loss": 0.0059, | |
| "step": 2350 | |
| }, | |
| { | |
| "epoch": 11.08, | |
| "learning_rate": 8.892018779342724e-06, | |
| "loss": 0.0153, | |
| "step": 2360 | |
| }, | |
| { | |
| "epoch": 11.13, | |
| "learning_rate": 8.887323943661973e-06, | |
| "loss": 0.0057, | |
| "step": 2370 | |
| }, | |
| { | |
| "epoch": 11.17, | |
| "learning_rate": 8.882629107981222e-06, | |
| "loss": 0.0048, | |
| "step": 2380 | |
| }, | |
| { | |
| "epoch": 11.22, | |
| "learning_rate": 8.87793427230047e-06, | |
| "loss": 0.0118, | |
| "step": 2390 | |
| }, | |
| { | |
| "epoch": 11.27, | |
| "learning_rate": 8.87323943661972e-06, | |
| "loss": 0.0071, | |
| "step": 2400 | |
| }, | |
| { | |
| "epoch": 11.31, | |
| "learning_rate": 8.868544600938968e-06, | |
| "loss": 0.0098, | |
| "step": 2410 | |
| }, | |
| { | |
| "epoch": 11.36, | |
| "learning_rate": 8.863849765258217e-06, | |
| "loss": 0.0065, | |
| "step": 2420 | |
| }, | |
| { | |
| "epoch": 11.41, | |
| "learning_rate": 8.859154929577466e-06, | |
| "loss": 0.0051, | |
| "step": 2430 | |
| }, | |
| { | |
| "epoch": 11.46, | |
| "learning_rate": 8.854460093896714e-06, | |
| "loss": 0.0053, | |
| "step": 2440 | |
| }, | |
| { | |
| "epoch": 11.5, | |
| "learning_rate": 8.849765258215963e-06, | |
| "loss": 0.0024, | |
| "step": 2450 | |
| }, | |
| { | |
| "epoch": 11.55, | |
| "learning_rate": 8.845070422535212e-06, | |
| "loss": 0.0045, | |
| "step": 2460 | |
| }, | |
| { | |
| "epoch": 11.6, | |
| "learning_rate": 8.84037558685446e-06, | |
| "loss": 0.0172, | |
| "step": 2470 | |
| }, | |
| { | |
| "epoch": 11.64, | |
| "learning_rate": 8.83568075117371e-06, | |
| "loss": 0.0053, | |
| "step": 2480 | |
| }, | |
| { | |
| "epoch": 11.69, | |
| "learning_rate": 8.830985915492958e-06, | |
| "loss": 0.014, | |
| "step": 2490 | |
| }, | |
| { | |
| "epoch": 11.74, | |
| "learning_rate": 8.826291079812207e-06, | |
| "loss": 0.0178, | |
| "step": 2500 | |
| }, | |
| { | |
| "epoch": 11.78, | |
| "learning_rate": 8.821596244131456e-06, | |
| "loss": 0.0071, | |
| "step": 2510 | |
| }, | |
| { | |
| "epoch": 11.83, | |
| "learning_rate": 8.816901408450705e-06, | |
| "loss": 0.0086, | |
| "step": 2520 | |
| }, | |
| { | |
| "epoch": 11.88, | |
| "learning_rate": 8.812206572769954e-06, | |
| "loss": 0.0056, | |
| "step": 2530 | |
| }, | |
| { | |
| "epoch": 11.92, | |
| "learning_rate": 8.807511737089202e-06, | |
| "loss": 0.0029, | |
| "step": 2540 | |
| }, | |
| { | |
| "epoch": 11.97, | |
| "learning_rate": 8.802816901408451e-06, | |
| "loss": 0.0076, | |
| "step": 2550 | |
| }, | |
| { | |
| "epoch": 12.0, | |
| "eval_corporation_f1": 0.2708333333333333, | |
| "eval_creative-work_f1": 0.39852398523985244, | |
| "eval_group_f1": 0.33333333333333337, | |
| "eval_location_f1": 0.6740331491712707, | |
| "eval_loss": 0.36174312233924866, | |
| "eval_overall_accuracy": 0.9473903221971228, | |
| "eval_overall_f1": 0.6143187066974597, | |
| "eval_overall_precision": 0.6919875130072841, | |
| "eval_overall_recall": 0.5523255813953488, | |
| "eval_person_f1": 0.75658419792498, | |
| "eval_product_f1": 0.3524590163934426, | |
| "eval_runtime": 5.9165, | |
| "eval_samples_per_second": 170.539, | |
| "eval_steps_per_second": 10.817, | |
| "step": 2556 | |
| }, | |
| { | |
| "epoch": 12.02, | |
| "learning_rate": 8.7981220657277e-06, | |
| "loss": 0.0034, | |
| "step": 2560 | |
| }, | |
| { | |
| "epoch": 12.07, | |
| "learning_rate": 8.793427230046949e-06, | |
| "loss": 0.0029, | |
| "step": 2570 | |
| }, | |
| { | |
| "epoch": 12.11, | |
| "learning_rate": 8.7887323943662e-06, | |
| "loss": 0.0084, | |
| "step": 2580 | |
| }, | |
| { | |
| "epoch": 12.16, | |
| "learning_rate": 8.784037558685446e-06, | |
| "loss": 0.0058, | |
| "step": 2590 | |
| }, | |
| { | |
| "epoch": 12.21, | |
| "learning_rate": 8.779342723004695e-06, | |
| "loss": 0.0025, | |
| "step": 2600 | |
| }, | |
| { | |
| "epoch": 12.25, | |
| "learning_rate": 8.774647887323946e-06, | |
| "loss": 0.0094, | |
| "step": 2610 | |
| }, | |
| { | |
| "epoch": 12.3, | |
| "learning_rate": 8.769953051643193e-06, | |
| "loss": 0.0051, | |
| "step": 2620 | |
| }, | |
| { | |
| "epoch": 12.35, | |
| "learning_rate": 8.765258215962442e-06, | |
| "loss": 0.0064, | |
| "step": 2630 | |
| }, | |
| { | |
| "epoch": 12.39, | |
| "learning_rate": 8.760563380281692e-06, | |
| "loss": 0.0095, | |
| "step": 2640 | |
| }, | |
| { | |
| "epoch": 12.44, | |
| "learning_rate": 8.755868544600939e-06, | |
| "loss": 0.0049, | |
| "step": 2650 | |
| }, | |
| { | |
| "epoch": 12.49, | |
| "learning_rate": 8.751173708920188e-06, | |
| "loss": 0.0114, | |
| "step": 2660 | |
| }, | |
| { | |
| "epoch": 12.54, | |
| "learning_rate": 8.746478873239437e-06, | |
| "loss": 0.0054, | |
| "step": 2670 | |
| }, | |
| { | |
| "epoch": 12.58, | |
| "learning_rate": 8.741784037558685e-06, | |
| "loss": 0.0108, | |
| "step": 2680 | |
| }, | |
| { | |
| "epoch": 12.63, | |
| "learning_rate": 8.737089201877936e-06, | |
| "loss": 0.0052, | |
| "step": 2690 | |
| }, | |
| { | |
| "epoch": 12.68, | |
| "learning_rate": 8.732394366197183e-06, | |
| "loss": 0.0086, | |
| "step": 2700 | |
| }, | |
| { | |
| "epoch": 12.72, | |
| "learning_rate": 8.727699530516432e-06, | |
| "loss": 0.0019, | |
| "step": 2710 | |
| }, | |
| { | |
| "epoch": 12.77, | |
| "learning_rate": 8.723004694835682e-06, | |
| "loss": 0.0024, | |
| "step": 2720 | |
| }, | |
| { | |
| "epoch": 12.82, | |
| "learning_rate": 8.71830985915493e-06, | |
| "loss": 0.0038, | |
| "step": 2730 | |
| }, | |
| { | |
| "epoch": 12.86, | |
| "learning_rate": 8.713615023474178e-06, | |
| "loss": 0.0097, | |
| "step": 2740 | |
| }, | |
| { | |
| "epoch": 12.91, | |
| "learning_rate": 8.708920187793429e-06, | |
| "loss": 0.0054, | |
| "step": 2750 | |
| }, | |
| { | |
| "epoch": 12.96, | |
| "learning_rate": 8.704225352112676e-06, | |
| "loss": 0.0042, | |
| "step": 2760 | |
| }, | |
| { | |
| "epoch": 13.0, | |
| "eval_corporation_f1": 0.24778761061946902, | |
| "eval_creative-work_f1": 0.39148936170212767, | |
| "eval_group_f1": 0.35211267605633806, | |
| "eval_location_f1": 0.656084656084656, | |
| "eval_loss": 0.3746570348739624, | |
| "eval_overall_accuracy": 0.9472714302698847, | |
| "eval_overall_f1": 0.6219790241678067, | |
| "eval_overall_precision": 0.6895854398382204, | |
| "eval_overall_recall": 0.5664451827242525, | |
| "eval_person_f1": 0.7741935483870966, | |
| "eval_product_f1": 0.35390946502057613, | |
| "eval_runtime": 6.4418, | |
| "eval_samples_per_second": 156.633, | |
| "eval_steps_per_second": 9.935, | |
| "step": 2769 | |
| }, | |
| { | |
| "epoch": 13.0, | |
| "learning_rate": 8.699530516431926e-06, | |
| "loss": 0.0039, | |
| "step": 2770 | |
| }, | |
| { | |
| "epoch": 13.05, | |
| "learning_rate": 8.694835680751175e-06, | |
| "loss": 0.0017, | |
| "step": 2780 | |
| }, | |
| { | |
| "epoch": 13.1, | |
| "learning_rate": 8.690140845070422e-06, | |
| "loss": 0.0102, | |
| "step": 2790 | |
| }, | |
| { | |
| "epoch": 13.15, | |
| "learning_rate": 8.685446009389673e-06, | |
| "loss": 0.0013, | |
| "step": 2800 | |
| }, | |
| { | |
| "epoch": 13.19, | |
| "learning_rate": 8.680751173708921e-06, | |
| "loss": 0.0076, | |
| "step": 2810 | |
| }, | |
| { | |
| "epoch": 13.24, | |
| "learning_rate": 8.676056338028169e-06, | |
| "loss": 0.0108, | |
| "step": 2820 | |
| }, | |
| { | |
| "epoch": 13.29, | |
| "learning_rate": 8.671361502347419e-06, | |
| "loss": 0.0107, | |
| "step": 2830 | |
| }, | |
| { | |
| "epoch": 13.33, | |
| "learning_rate": 8.666666666666668e-06, | |
| "loss": 0.0058, | |
| "step": 2840 | |
| }, | |
| { | |
| "epoch": 13.38, | |
| "learning_rate": 8.661971830985915e-06, | |
| "loss": 0.0033, | |
| "step": 2850 | |
| }, | |
| { | |
| "epoch": 13.43, | |
| "learning_rate": 8.657276995305165e-06, | |
| "loss": 0.0028, | |
| "step": 2860 | |
| }, | |
| { | |
| "epoch": 13.47, | |
| "learning_rate": 8.652582159624414e-06, | |
| "loss": 0.0046, | |
| "step": 2870 | |
| }, | |
| { | |
| "epoch": 13.52, | |
| "learning_rate": 8.647887323943663e-06, | |
| "loss": 0.002, | |
| "step": 2880 | |
| }, | |
| { | |
| "epoch": 13.57, | |
| "learning_rate": 8.643192488262912e-06, | |
| "loss": 0.0042, | |
| "step": 2890 | |
| }, | |
| { | |
| "epoch": 13.62, | |
| "learning_rate": 8.63849765258216e-06, | |
| "loss": 0.0066, | |
| "step": 2900 | |
| }, | |
| { | |
| "epoch": 13.66, | |
| "learning_rate": 8.63380281690141e-06, | |
| "loss": 0.0022, | |
| "step": 2910 | |
| }, | |
| { | |
| "epoch": 13.71, | |
| "learning_rate": 8.629107981220658e-06, | |
| "loss": 0.008, | |
| "step": 2920 | |
| }, | |
| { | |
| "epoch": 13.76, | |
| "learning_rate": 8.624413145539907e-06, | |
| "loss": 0.0037, | |
| "step": 2930 | |
| }, | |
| { | |
| "epoch": 13.8, | |
| "learning_rate": 8.619718309859156e-06, | |
| "loss": 0.0106, | |
| "step": 2940 | |
| }, | |
| { | |
| "epoch": 13.85, | |
| "learning_rate": 8.615023474178405e-06, | |
| "loss": 0.0031, | |
| "step": 2950 | |
| }, | |
| { | |
| "epoch": 13.9, | |
| "learning_rate": 8.610328638497653e-06, | |
| "loss": 0.0023, | |
| "step": 2960 | |
| }, | |
| { | |
| "epoch": 13.94, | |
| "learning_rate": 8.605633802816902e-06, | |
| "loss": 0.0118, | |
| "step": 2970 | |
| }, | |
| { | |
| "epoch": 13.99, | |
| "learning_rate": 8.600938967136151e-06, | |
| "loss": 0.0049, | |
| "step": 2980 | |
| }, | |
| { | |
| "epoch": 14.0, | |
| "eval_corporation_f1": 0.29752066115702475, | |
| "eval_creative-work_f1": 0.44360902255639095, | |
| "eval_group_f1": 0.36241610738255037, | |
| "eval_location_f1": 0.6834170854271356, | |
| "eval_loss": 0.3375568985939026, | |
| "eval_overall_accuracy": 0.950897634050648, | |
| "eval_overall_f1": 0.6430135786246168, | |
| "eval_overall_precision": 0.680259499536608, | |
| "eval_overall_recall": 0.6096345514950167, | |
| "eval_person_f1": 0.7901614142966948, | |
| "eval_product_f1": 0.388663967611336, | |
| "eval_runtime": 6.509, | |
| "eval_samples_per_second": 155.016, | |
| "eval_steps_per_second": 9.833, | |
| "step": 2982 | |
| }, | |
| { | |
| "epoch": 14.0, | |
| "step": 2982, | |
| "total_flos": 1093917417133104.0, | |
| "train_loss": 0.01162932096780006, | |
| "train_runtime": 800.644, | |
| "train_samples_per_second": 423.909, | |
| "train_steps_per_second": 26.604 | |
| } | |
| ], | |
| "max_steps": 21300, | |
| "num_train_epochs": 100, | |
| "total_flos": 1093917417133104.0, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |