| { | |
| "best_metric": 0.7434775214630398, | |
| "best_model_checkpoint": "tiny-llama-lora-new/checkpoint-2400", | |
| "epoch": 14.930015552099533, | |
| "eval_steps": 500, | |
| "global_step": 2400, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 1.0, | |
| "eval_accuracy": 0.8001549186676995, | |
| "eval_f1_macro": 0.6790360166213908, | |
| "eval_f1_micro": 0.8001549186676995, | |
| "eval_f1_weighted": 0.7959404181222782, | |
| "eval_loss": 0.6614832282066345, | |
| "eval_macro_fpr": 0.01824293967244035, | |
| "eval_macro_sensitivity": 0.6677834011062839, | |
| "eval_macro_specificity": 0.9848478627893485, | |
| "eval_precision": 0.8040294558027004, | |
| "eval_precision_macro": 0.7265846774862527, | |
| "eval_recall": 0.8001549186676995, | |
| "eval_recall_macro": 0.6677834011062839, | |
| "eval_runtime": 225.9328, | |
| "eval_samples_per_second": 5.714, | |
| "eval_steps_per_second": 0.717, | |
| "eval_weighted_fpr": 0.017527173913043478, | |
| "eval_weighted_sensitivity": 0.8001549186676995, | |
| "eval_weighted_specificity": 0.97256302317253, | |
| "step": 160 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_accuracy": 0.8063516653756778, | |
| "eval_f1_macro": 0.723523893027765, | |
| "eval_f1_micro": 0.8063516653756778, | |
| "eval_f1_weighted": 0.8039099652383133, | |
| "eval_loss": 0.6995529532432556, | |
| "eval_macro_fpr": 0.01771710819775641, | |
| "eval_macro_sensitivity": 0.7207263437235008, | |
| "eval_macro_specificity": 0.9853387258716434, | |
| "eval_precision": 0.8110438944279544, | |
| "eval_precision_macro": 0.7448100337436813, | |
| "eval_recall": 0.8063516653756778, | |
| "eval_recall_macro": 0.7207263437235008, | |
| "eval_runtime": 230.7469, | |
| "eval_samples_per_second": 5.595, | |
| "eval_steps_per_second": 0.702, | |
| "eval_weighted_fpr": 0.01686454398273071, | |
| "eval_weighted_sensitivity": 0.8063516653756778, | |
| "eval_weighted_specificity": 0.9737292226989717, | |
| "step": 321 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_accuracy": 0.8125484120836561, | |
| "eval_f1_macro": 0.7179663090811256, | |
| "eval_f1_micro": 0.8125484120836561, | |
| "eval_f1_weighted": 0.8085118934271095, | |
| "eval_loss": 0.8202366828918457, | |
| "eval_macro_fpr": 0.017097652587672894, | |
| "eval_macro_sensitivity": 0.7080107913825292, | |
| "eval_macro_specificity": 0.985578595986596, | |
| "eval_precision": 0.8118501609528374, | |
| "eval_precision_macro": 0.7576540712122397, | |
| "eval_recall": 0.8125484120836561, | |
| "eval_recall_macro": 0.7080107913825292, | |
| "eval_runtime": 231.655, | |
| "eval_samples_per_second": 5.573, | |
| "eval_steps_per_second": 0.699, | |
| "eval_weighted_fpr": 0.016211146838156484, | |
| "eval_weighted_sensitivity": 0.8125484120836561, | |
| "eval_weighted_specificity": 0.9711305277152844, | |
| "step": 482 | |
| }, | |
| { | |
| "epoch": 3.11, | |
| "learning_rate": 3.958333333333333e-05, | |
| "loss": 0.2932, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "eval_accuracy": 0.8140975987606507, | |
| "eval_f1_macro": 0.7415140921427528, | |
| "eval_f1_micro": 0.8140975987606507, | |
| "eval_f1_weighted": 0.8153735617531811, | |
| "eval_loss": 0.9493006467819214, | |
| "eval_macro_fpr": 0.016573672675540378, | |
| "eval_macro_sensitivity": 0.7326751453453961, | |
| "eval_macro_specificity": 0.9859001747376132, | |
| "eval_precision": 0.8204052204072106, | |
| "eval_precision_macro": 0.759251934556511, | |
| "eval_recall": 0.8140975987606507, | |
| "eval_recall_macro": 0.7326751453453961, | |
| "eval_runtime": 230.2371, | |
| "eval_samples_per_second": 5.607, | |
| "eval_steps_per_second": 0.704, | |
| "eval_weighted_fpr": 0.016049217600641968, | |
| "eval_weighted_sensitivity": 0.8140975987606507, | |
| "eval_weighted_specificity": 0.974405022303545, | |
| "step": 643 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "eval_accuracy": 0.8109992254066615, | |
| "eval_f1_macro": 0.7413292975124365, | |
| "eval_f1_micro": 0.8109992254066615, | |
| "eval_f1_weighted": 0.8086569922767037, | |
| "eval_loss": 1.0610005855560303, | |
| "eval_macro_fpr": 0.017150663222076427, | |
| "eval_macro_sensitivity": 0.7427303328596474, | |
| "eval_macro_specificity": 0.9856545654845559, | |
| "eval_precision": 0.8110273755487133, | |
| "eval_precision_macro": 0.7595846993577862, | |
| "eval_recall": 0.8109992254066615, | |
| "eval_recall_macro": 0.7427303328596474, | |
| "eval_runtime": 232.3513, | |
| "eval_samples_per_second": 5.556, | |
| "eval_steps_per_second": 0.697, | |
| "eval_weighted_fpr": 0.016373641121997046, | |
| "eval_weighted_sensitivity": 0.8109992254066615, | |
| "eval_weighted_specificity": 0.9738192568616761, | |
| "step": 803 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "eval_accuracy": 0.814872192099148, | |
| "eval_f1_macro": 0.740834009056896, | |
| "eval_f1_micro": 0.814872192099148, | |
| "eval_f1_weighted": 0.812828873292886, | |
| "eval_loss": 1.1361573934555054, | |
| "eval_macro_fpr": 0.016745822161733162, | |
| "eval_macro_sensitivity": 0.7380044964762863, | |
| "eval_macro_specificity": 0.9859281636821422, | |
| "eval_precision": 0.8159522433901665, | |
| "eval_precision_macro": 0.7731408058998965, | |
| "eval_recall": 0.814872192099148, | |
| "eval_recall_macro": 0.7380044964762863, | |
| "eval_runtime": 230.3919, | |
| "eval_samples_per_second": 5.603, | |
| "eval_steps_per_second": 0.703, | |
| "eval_weighted_fpr": 0.015968463954032203, | |
| "eval_weighted_sensitivity": 0.814872192099148, | |
| "eval_weighted_specificity": 0.9740502631329879, | |
| "step": 964 | |
| }, | |
| { | |
| "epoch": 6.22, | |
| "learning_rate": 2.916666666666667e-05, | |
| "loss": 0.0107, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "eval_accuracy": 0.8102246320681642, | |
| "eval_f1_macro": 0.734324336127792, | |
| "eval_f1_micro": 0.8102246320681642, | |
| "eval_f1_weighted": 0.8084846815925374, | |
| "eval_loss": 1.1712960004806519, | |
| "eval_macro_fpr": 0.017129374121876983, | |
| "eval_macro_sensitivity": 0.730990149259248, | |
| "eval_macro_specificity": 0.9855872802214631, | |
| "eval_precision": 0.812320444346659, | |
| "eval_precision_macro": 0.7734170589513789, | |
| "eval_recall": 0.8102246320681642, | |
| "eval_recall_macro": 0.730990149259248, | |
| "eval_runtime": 232.06, | |
| "eval_samples_per_second": 5.563, | |
| "eval_steps_per_second": 0.698, | |
| "eval_weighted_fpr": 0.016455101081335213, | |
| "eval_weighted_sensitivity": 0.8102246320681642, | |
| "eval_weighted_specificity": 0.973584571253782, | |
| "step": 1125 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "eval_accuracy": 0.8156467854376452, | |
| "eval_f1_macro": 0.7374069155892726, | |
| "eval_f1_micro": 0.8156467854376452, | |
| "eval_f1_weighted": 0.8127779657126307, | |
| "eval_loss": 1.1785770654678345, | |
| "eval_macro_fpr": 0.016642226736202675, | |
| "eval_macro_sensitivity": 0.7348966803697986, | |
| "eval_macro_specificity": 0.9859780376275717, | |
| "eval_precision": 0.8140580469506681, | |
| "eval_precision_macro": 0.7655811603242806, | |
| "eval_recall": 0.8156467854376452, | |
| "eval_recall_macro": 0.7348966803697986, | |
| "eval_runtime": 230.6812, | |
| "eval_samples_per_second": 5.596, | |
| "eval_steps_per_second": 0.702, | |
| "eval_weighted_fpr": 0.01588785046728972, | |
| "eval_weighted_sensitivity": 0.8156467854376452, | |
| "eval_weighted_specificity": 0.9740237789759322, | |
| "step": 1286 | |
| }, | |
| { | |
| "epoch": 9.0, | |
| "eval_accuracy": 0.8187451587916343, | |
| "eval_f1_macro": 0.7400115374685853, | |
| "eval_f1_micro": 0.8187451587916342, | |
| "eval_f1_weighted": 0.8156936902580679, | |
| "eval_loss": 1.1960150003433228, | |
| "eval_macro_fpr": 0.01632755569207043, | |
| "eval_macro_sensitivity": 0.7367529624892385, | |
| "eval_macro_specificity": 0.9862031650054388, | |
| "eval_precision": 0.8170206925865887, | |
| "eval_precision_macro": 0.769266125041975, | |
| "eval_recall": 0.8187451587916343, | |
| "eval_recall_macro": 0.7367529624892385, | |
| "eval_runtime": 231.8038, | |
| "eval_samples_per_second": 5.569, | |
| "eval_steps_per_second": 0.699, | |
| "eval_weighted_fpr": 0.015566790846194785, | |
| "eval_weighted_sensitivity": 0.8187451587916343, | |
| "eval_weighted_specificity": 0.9743023162899509, | |
| "step": 1446 | |
| }, | |
| { | |
| "epoch": 9.33, | |
| "learning_rate": 1.8750000000000002e-05, | |
| "loss": 0.0016, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 10.0, | |
| "eval_accuracy": 0.8156467854376452, | |
| "eval_f1_macro": 0.7376161090278297, | |
| "eval_f1_micro": 0.8156467854376452, | |
| "eval_f1_weighted": 0.8130986313804958, | |
| "eval_loss": 1.204884648323059, | |
| "eval_macro_fpr": 0.016621895484736447, | |
| "eval_macro_sensitivity": 0.7352544079731995, | |
| "eval_macro_specificity": 0.9859824832056153, | |
| "eval_precision": 0.8149926729426089, | |
| "eval_precision_macro": 0.765906061186384, | |
| "eval_recall": 0.8156467854376452, | |
| "eval_recall_macro": 0.7352544079731995, | |
| "eval_runtime": 230.9138, | |
| "eval_samples_per_second": 5.591, | |
| "eval_steps_per_second": 0.702, | |
| "eval_weighted_fpr": 0.01588785046728972, | |
| "eval_weighted_sensitivity": 0.8156467854376452, | |
| "eval_weighted_specificity": 0.9740904626465837, | |
| "step": 1607 | |
| }, | |
| { | |
| "epoch": 11.0, | |
| "eval_accuracy": 0.8156467854376452, | |
| "eval_f1_macro": 0.7376791177200052, | |
| "eval_f1_micro": 0.8156467854376452, | |
| "eval_f1_weighted": 0.8129515156545265, | |
| "eval_loss": 1.2137137651443481, | |
| "eval_macro_fpr": 0.016631267825388357, | |
| "eval_macro_sensitivity": 0.7352648717267571, | |
| "eval_macro_specificity": 0.9859807497677747, | |
| "eval_precision": 0.81468290664357, | |
| "eval_precision_macro": 0.766075737607577, | |
| "eval_recall": 0.8156467854376452, | |
| "eval_recall_macro": 0.7352648717267571, | |
| "eval_runtime": 234.0704, | |
| "eval_samples_per_second": 5.515, | |
| "eval_steps_per_second": 0.692, | |
| "eval_weighted_fpr": 0.01588785046728972, | |
| "eval_weighted_sensitivity": 0.8156467854376452, | |
| "eval_weighted_specificity": 0.9740644610789766, | |
| "step": 1768 | |
| }, | |
| { | |
| "epoch": 12.0, | |
| "eval_accuracy": 0.8156467854376452, | |
| "eval_f1_macro": 0.7378911666248646, | |
| "eval_f1_micro": 0.8156467854376452, | |
| "eval_f1_weighted": 0.8128626681643639, | |
| "eval_loss": 1.2158095836639404, | |
| "eval_macro_fpr": 0.01664811744097862, | |
| "eval_macro_sensitivity": 0.7352648717267571, | |
| "eval_macro_specificity": 0.9859669931344103, | |
| "eval_precision": 0.8144771699978494, | |
| "eval_precision_macro": 0.7664387794391382, | |
| "eval_recall": 0.8156467854376452, | |
| "eval_recall_macro": 0.7352648717267571, | |
| "eval_runtime": 231.833, | |
| "eval_samples_per_second": 5.569, | |
| "eval_steps_per_second": 0.699, | |
| "eval_weighted_fpr": 0.01588785046728972, | |
| "eval_weighted_sensitivity": 0.8156467854376452, | |
| "eval_weighted_specificity": 0.9738581115785108, | |
| "step": 1929 | |
| }, | |
| { | |
| "epoch": 12.44, | |
| "learning_rate": 8.333333333333334e-06, | |
| "loss": 0.0011, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 13.0, | |
| "eval_accuracy": 0.8187451587916343, | |
| "eval_f1_macro": 0.7425477382056641, | |
| "eval_f1_micro": 0.8187451587916342, | |
| "eval_f1_weighted": 0.8157578026261842, | |
| "eval_loss": 1.2201900482177734, | |
| "eval_macro_fpr": 0.01633145053296977, | |
| "eval_macro_sensitivity": 0.7372241049428965, | |
| "eval_macro_specificity": 0.9861894646777546, | |
| "eval_precision": 0.8168736065801445, | |
| "eval_precision_macro": 0.7720479168106198, | |
| "eval_recall": 0.8187451587916343, | |
| "eval_recall_macro": 0.7372241049428965, | |
| "eval_runtime": 230.3364, | |
| "eval_samples_per_second": 5.605, | |
| "eval_steps_per_second": 0.703, | |
| "eval_weighted_fpr": 0.015566790846194785, | |
| "eval_weighted_sensitivity": 0.8187451587916343, | |
| "eval_weighted_specificity": 0.974096811374682, | |
| "step": 2089 | |
| }, | |
| { | |
| "epoch": 14.0, | |
| "eval_accuracy": 0.8187451587916343, | |
| "eval_f1_macro": 0.7425477382056641, | |
| "eval_f1_micro": 0.8187451587916342, | |
| "eval_f1_weighted": 0.8157578026261842, | |
| "eval_loss": 1.2229193449020386, | |
| "eval_macro_fpr": 0.01633145053296977, | |
| "eval_macro_sensitivity": 0.7372241049428965, | |
| "eval_macro_specificity": 0.9861894646777546, | |
| "eval_precision": 0.8168736065801445, | |
| "eval_precision_macro": 0.7720479168106198, | |
| "eval_recall": 0.8187451587916343, | |
| "eval_recall_macro": 0.7372241049428965, | |
| "eval_runtime": 234.1844, | |
| "eval_samples_per_second": 5.513, | |
| "eval_steps_per_second": 0.692, | |
| "eval_weighted_fpr": 0.015566790846194785, | |
| "eval_weighted_sensitivity": 0.8187451587916343, | |
| "eval_weighted_specificity": 0.974096811374682, | |
| "step": 2250 | |
| }, | |
| { | |
| "epoch": 14.93, | |
| "eval_accuracy": 0.820294345468629, | |
| "eval_f1_macro": 0.7434775214630398, | |
| "eval_f1_micro": 0.8202943454686291, | |
| "eval_f1_weighted": 0.8172740569769142, | |
| "eval_loss": 1.2251933813095093, | |
| "eval_macro_fpr": 0.01617133673495677, | |
| "eval_macro_sensitivity": 0.7379816807004722, | |
| "eval_macro_specificity": 0.9863045756599175, | |
| "eval_precision": 0.8184108267929948, | |
| "eval_precision_macro": 0.77323629548042, | |
| "eval_recall": 0.820294345468629, | |
| "eval_recall_macro": 0.7379816807004722, | |
| "eval_runtime": 286.1496, | |
| "eval_samples_per_second": 4.512, | |
| "eval_steps_per_second": 0.566, | |
| "eval_weighted_fpr": 0.015407092575375215, | |
| "eval_weighted_sensitivity": 0.820294345468629, | |
| "eval_weighted_specificity": 0.9742742894301325, | |
| "step": 2400 | |
| } | |
| ], | |
| "logging_steps": 500, | |
| "max_steps": 2400, | |
| "num_train_epochs": 15, | |
| "save_steps": 500, | |
| "total_flos": 2.313478048139182e+17, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |