cares-bert-base / trainer_state.json
chizhik
re-trained model after eliminating annotation errors
2a0b29c
{
"best_metric": 0.8249799337347952,
"best_model_checkpoint": "./CARES/checkpoints/bert-ba-stratified/run-9/checkpoint-3976",
"epoch": 56.0,
"global_step": 3976,
"is_hyper_param_search": true,
"is_local_process_zero": true,
"is_world_process_zero": true,
"log_history": [
{
"epoch": 2.0,
"eval_loss": 0.24943208694458008,
"eval_macro_f1": 0.08911588063155029,
"eval_macro_precision": 0.1020147123407993,
"eval_macro_recall": 0.08017933766090879,
"eval_micro_f1": 0.27627302275189597,
"eval_micro_precision": 0.7750759878419453,
"eval_micro_recall": 0.16809492419248517,
"eval_runtime": 2.9775,
"eval_samples_per_second": 324.431,
"eval_steps_per_second": 20.487,
"step": 142
},
{
"epoch": 4.0,
"eval_loss": 0.1698431819677353,
"eval_macro_f1": 0.30238618603356054,
"eval_macro_precision": 0.4320649543604988,
"eval_macro_recall": 0.2644492382273683,
"eval_micro_f1": 0.6521739130434782,
"eval_micro_precision": 0.8376421923474664,
"eval_micro_recall": 0.5339485827290705,
"eval_runtime": 2.979,
"eval_samples_per_second": 324.273,
"eval_steps_per_second": 20.477,
"step": 284
},
{
"epoch": 6.0,
"eval_loss": 0.14047418534755707,
"eval_macro_f1": 0.40223830727079396,
"eval_macro_precision": 0.4784721691006365,
"eval_macro_recall": 0.3687052580973402,
"eval_micro_f1": 0.739880059970015,
"eval_micro_precision": 0.8575152041702867,
"eval_micro_recall": 0.6506262359920897,
"eval_runtime": 2.9805,
"eval_samples_per_second": 324.11,
"eval_steps_per_second": 20.467,
"step": 426
},
{
"epoch": 7.04,
"learning_rate": 3.279287453609026e-05,
"loss": 0.2244,
"step": 500
},
{
"epoch": 8.0,
"eval_loss": 0.1187577173113823,
"eval_macro_f1": 0.5347249220382584,
"eval_macro_precision": 0.7327975628305986,
"eval_macro_recall": 0.4754068941604508,
"eval_micro_f1": 0.7880299251870324,
"eval_micro_precision": 0.8573643410852713,
"eval_micro_recall": 0.7290705339485827,
"eval_runtime": 2.9805,
"eval_samples_per_second": 324.108,
"eval_steps_per_second": 20.466,
"step": 568
},
{
"epoch": 10.0,
"eval_loss": 0.11074026674032211,
"eval_macro_f1": 0.5991127961103198,
"eval_macro_precision": 0.6991665233729463,
"eval_macro_recall": 0.5596673950826421,
"eval_micro_f1": 0.8128196385952949,
"eval_micro_precision": 0.8418079096045198,
"eval_micro_recall": 0.7857613711272248,
"eval_runtime": 2.9796,
"eval_samples_per_second": 324.206,
"eval_steps_per_second": 20.473,
"step": 710
},
{
"epoch": 12.0,
"eval_loss": 0.10078531503677368,
"eval_macro_f1": 0.6568773778388772,
"eval_macro_precision": 0.700413763066982,
"eval_macro_recall": 0.6283428000904666,
"eval_micro_f1": 0.8347529812606473,
"eval_micro_precision": 0.8638928067700987,
"eval_micro_recall": 0.8075148319050758,
"eval_runtime": 2.979,
"eval_samples_per_second": 324.274,
"eval_steps_per_second": 20.477,
"step": 852
},
{
"epoch": 14.0,
"eval_loss": 0.10221733897924423,
"eval_macro_f1": 0.6758575377881516,
"eval_macro_precision": 0.838910780569426,
"eval_macro_recall": 0.6238243641780066,
"eval_micro_f1": 0.8337912087912088,
"eval_micro_precision": 0.8702508960573476,
"eval_micro_recall": 0.8002636783124588,
"eval_runtime": 2.9795,
"eval_samples_per_second": 324.21,
"eval_steps_per_second": 20.473,
"step": 994
},
{
"epoch": 14.08,
"learning_rate": 3.0308565859113728e-05,
"loss": 0.0513,
"step": 1000
},
{
"epoch": 16.0,
"eval_loss": 0.10313227772712708,
"eval_macro_f1": 0.7298003500123689,
"eval_macro_precision": 0.8118751777216693,
"eval_macro_recall": 0.6944187826598622,
"eval_micro_f1": 0.8354006034193765,
"eval_micro_precision": 0.849931787175989,
"eval_micro_recall": 0.8213579433091628,
"eval_runtime": 2.9811,
"eval_samples_per_second": 324.042,
"eval_steps_per_second": 20.462,
"step": 1136
},
{
"epoch": 18.0,
"eval_loss": 0.10032625496387482,
"eval_macro_f1": 0.7681297243773157,
"eval_macro_precision": 0.8958394795684446,
"eval_macro_recall": 0.7124194145895655,
"eval_micro_f1": 0.8443093549476527,
"eval_micro_precision": 0.8656509695290858,
"eval_micro_recall": 0.8239947264337508,
"eval_runtime": 2.9813,
"eval_samples_per_second": 324.018,
"eval_steps_per_second": 20.461,
"step": 1278
},
{
"epoch": 20.0,
"eval_loss": 0.1044757142663002,
"eval_macro_f1": 0.7829375079467304,
"eval_macro_precision": 0.8858384241353942,
"eval_macro_recall": 0.7367403001149204,
"eval_micro_f1": 0.8420698924731183,
"eval_micro_precision": 0.8588074023303632,
"eval_micro_recall": 0.8259723137771918,
"eval_runtime": 2.9861,
"eval_samples_per_second": 323.499,
"eval_steps_per_second": 20.428,
"step": 1420
},
{
"epoch": 21.13,
"learning_rate": 2.7824257182137193e-05,
"loss": 0.0183,
"step": 1500
},
{
"epoch": 22.0,
"eval_loss": 0.1039622500538826,
"eval_macro_f1": 0.7902798824417182,
"eval_macro_precision": 0.89276275853935,
"eval_macro_recall": 0.7403232660636272,
"eval_micro_f1": 0.8479512360311547,
"eval_micro_precision": 0.871866295264624,
"eval_micro_recall": 0.8253131179960448,
"eval_runtime": 2.977,
"eval_samples_per_second": 324.483,
"eval_steps_per_second": 20.49,
"step": 1562
},
{
"epoch": 24.0,
"eval_loss": 0.10449391603469849,
"eval_macro_f1": 0.7973957256263516,
"eval_macro_precision": 0.8897797609904068,
"eval_macro_recall": 0.7482788788828134,
"eval_micro_f1": 0.8530297957817209,
"eval_micro_precision": 0.8666666666666667,
"eval_micro_recall": 0.8398154251812788,
"eval_runtime": 2.982,
"eval_samples_per_second": 323.947,
"eval_steps_per_second": 20.456,
"step": 1704
},
{
"epoch": 26.0,
"eval_loss": 0.10751193016767502,
"eval_macro_f1": 0.815099576727704,
"eval_macro_precision": 0.8596170620799346,
"eval_macro_recall": 0.7893940744657464,
"eval_micro_f1": 0.8517287234042553,
"eval_micro_precision": 0.8591549295774648,
"eval_micro_recall": 0.8444297956493079,
"eval_runtime": 2.9874,
"eval_samples_per_second": 323.362,
"eval_steps_per_second": 20.419,
"step": 1846
},
{
"epoch": 28.0,
"eval_loss": 0.10826986283063889,
"eval_macro_f1": 0.8122978188941863,
"eval_macro_precision": 0.8818206343623207,
"eval_macro_recall": 0.7768327155632533,
"eval_micro_f1": 0.8507362784471219,
"eval_micro_precision": 0.8640380693405847,
"eval_micro_recall": 0.8378378378378378,
"eval_runtime": 2.9805,
"eval_samples_per_second": 324.106,
"eval_steps_per_second": 20.466,
"step": 1988
},
{
"epoch": 28.17,
"learning_rate": 2.5339948505160657e-05,
"loss": 0.0094,
"step": 2000
},
{
"epoch": 30.0,
"eval_loss": 0.11172767728567123,
"eval_macro_f1": 0.7952069303877679,
"eval_macro_precision": 0.8834731091800803,
"eval_macro_recall": 0.7589521997225694,
"eval_micro_f1": 0.8510210913960494,
"eval_micro_precision": 0.8646258503401361,
"eval_micro_recall": 0.8378378378378378,
"eval_runtime": 2.9811,
"eval_samples_per_second": 324.039,
"eval_steps_per_second": 20.462,
"step": 2130
},
{
"epoch": 32.0,
"eval_loss": 0.12214264273643494,
"eval_macro_f1": 0.784963786280585,
"eval_macro_precision": 0.8399655989881138,
"eval_macro_recall": 0.7539663033116887,
"eval_micro_f1": 0.8395881766854865,
"eval_micro_precision": 0.8460508701472557,
"eval_micro_recall": 0.8332234673698088,
"eval_runtime": 2.9802,
"eval_samples_per_second": 324.138,
"eval_steps_per_second": 20.468,
"step": 2272
},
{
"epoch": 34.0,
"eval_loss": 0.10992709547281265,
"eval_macro_f1": 0.818123937895116,
"eval_macro_precision": 0.846307748666965,
"eval_macro_recall": 0.7959472696549352,
"eval_micro_f1": 0.8540829986613119,
"eval_micro_precision": 0.86743711760707,
"eval_micro_recall": 0.8411338167435728,
"eval_runtime": 2.9817,
"eval_samples_per_second": 323.976,
"eval_steps_per_second": 20.458,
"step": 2414
},
{
"epoch": 35.21,
"learning_rate": 2.2855639828184125e-05,
"loss": 0.0088,
"step": 2500
},
{
"epoch": 36.0,
"eval_loss": 0.11282340437173843,
"eval_macro_f1": 0.810588172484549,
"eval_macro_precision": 0.8881718400115531,
"eval_macro_recall": 0.7660828954423657,
"eval_micro_f1": 0.8533783783783784,
"eval_micro_precision": 0.8752598752598753,
"eval_micro_recall": 0.8325642715886619,
"eval_runtime": 2.9823,
"eval_samples_per_second": 323.913,
"eval_steps_per_second": 20.454,
"step": 2556
},
{
"epoch": 38.0,
"eval_loss": 0.11364943534135818,
"eval_macro_f1": 0.8086979072156089,
"eval_macro_precision": 0.874192444614366,
"eval_macro_recall": 0.7687841160082525,
"eval_micro_f1": 0.855510752688172,
"eval_micro_precision": 0.8725154215215901,
"eval_micro_recall": 0.8391562294001318,
"eval_runtime": 2.9856,
"eval_samples_per_second": 323.552,
"eval_steps_per_second": 20.431,
"step": 2698
},
{
"epoch": 40.0,
"eval_loss": 0.11377756297588348,
"eval_macro_f1": 0.8074550758048753,
"eval_macro_precision": 0.8601506214626871,
"eval_macro_recall": 0.7772364003410493,
"eval_micro_f1": 0.8550483172275909,
"eval_micro_precision": 0.8645552560646901,
"eval_micro_recall": 0.8457481872116018,
"eval_runtime": 2.987,
"eval_samples_per_second": 323.399,
"eval_steps_per_second": 20.422,
"step": 2840
},
{
"epoch": 42.0,
"eval_loss": 0.11456111818552017,
"eval_macro_f1": 0.8197602680311404,
"eval_macro_precision": 0.8652772327919502,
"eval_macro_recall": 0.7923572373451824,
"eval_micro_f1": 0.8584748584748586,
"eval_micro_precision": 0.8674293405114402,
"eval_micro_recall": 0.8497033618984838,
"eval_runtime": 2.9793,
"eval_samples_per_second": 324.237,
"eval_steps_per_second": 20.475,
"step": 2982
},
{
"epoch": 42.25,
"learning_rate": 2.0371331151207586e-05,
"loss": 0.0049,
"step": 3000
},
{
"epoch": 44.0,
"eval_loss": 0.11773423105478287,
"eval_macro_f1": 0.8086306686692242,
"eval_macro_precision": 0.8642281309223148,
"eval_macro_recall": 0.7764573931394076,
"eval_micro_f1": 0.8553291012362179,
"eval_micro_precision": 0.8672086720867209,
"eval_micro_recall": 0.8437705998681608,
"eval_runtime": 2.9813,
"eval_samples_per_second": 324.023,
"eval_steps_per_second": 20.461,
"step": 3124
},
{
"epoch": 46.0,
"eval_loss": 0.11780666559934616,
"eval_macro_f1": 0.8101599105843645,
"eval_macro_precision": 0.8691080726361069,
"eval_macro_recall": 0.7762766074908475,
"eval_micro_f1": 0.8565723793677205,
"eval_micro_precision": 0.8649193548387096,
"eval_micro_recall": 0.8483849703361899,
"eval_runtime": 2.9804,
"eval_samples_per_second": 324.123,
"eval_steps_per_second": 20.467,
"step": 3266
},
{
"epoch": 48.0,
"eval_loss": 0.11932362616062164,
"eval_macro_f1": 0.8165194442314286,
"eval_macro_precision": 0.8648115942931565,
"eval_macro_recall": 0.7864788866212832,
"eval_micro_f1": 0.8535201868535202,
"eval_micro_precision": 0.8641891891891892,
"eval_micro_recall": 0.8431114040870138,
"eval_runtime": 2.985,
"eval_samples_per_second": 323.616,
"eval_steps_per_second": 20.435,
"step": 3408
},
{
"epoch": 49.3,
"learning_rate": 1.7887022474231054e-05,
"loss": 0.0034,
"step": 3500
},
{
"epoch": 50.0,
"eval_loss": 0.12221735715866089,
"eval_macro_f1": 0.8141089592504215,
"eval_macro_precision": 0.8615361853719665,
"eval_macro_recall": 0.7844581807191786,
"eval_micro_f1": 0.8523714094856378,
"eval_micro_precision": 0.8639133378469871,
"eval_micro_recall": 0.8411338167435728,
"eval_runtime": 2.9803,
"eval_samples_per_second": 324.125,
"eval_steps_per_second": 20.468,
"step": 3550
},
{
"epoch": 52.0,
"eval_loss": 0.12201466411352158,
"eval_macro_f1": 0.8215839258353351,
"eval_macro_precision": 0.8618774470352207,
"eval_macro_recall": 0.795741448912531,
"eval_micro_f1": 0.8534223706176962,
"eval_micro_precision": 0.86468200270636,
"eval_micro_recall": 0.8424522083058669,
"eval_runtime": 2.9796,
"eval_samples_per_second": 324.208,
"eval_steps_per_second": 20.473,
"step": 3692
},
{
"epoch": 54.0,
"eval_loss": 0.12538054585456848,
"eval_macro_f1": 0.8134875725433206,
"eval_macro_precision": 0.8710944940417797,
"eval_macro_recall": 0.7763605326378977,
"eval_micro_f1": 0.85425782564793,
"eval_micro_precision": 0.8727647867950481,
"eval_micro_recall": 0.8365194462755439,
"eval_runtime": 2.9841,
"eval_samples_per_second": 323.717,
"eval_steps_per_second": 20.442,
"step": 3834
},
{
"epoch": 56.0,
"eval_loss": 0.12492760270833969,
"eval_macro_f1": 0.8249799337347952,
"eval_macro_precision": 0.8672879358583392,
"eval_macro_recall": 0.7974535529440212,
"eval_micro_f1": 0.8574290484140235,
"eval_micro_precision": 0.8687415426251691,
"eval_micro_recall": 0.8464073829927489,
"eval_runtime": 2.9821,
"eval_samples_per_second": 323.938,
"eval_steps_per_second": 20.456,
"step": 3976
}
],
"max_steps": 7100,
"num_train_epochs": 100,
"total_flos": 3.752304368656253e+16,
"trial_name": null,
"trial_params": {
"adam_epsilon": 2.4799103776060603e-09,
"learning_rate": 3.4268553890214325e-05,
"per_device_eval_batch_size": 16,
"per_device_train_batch_size": 32,
"seed": 326,
"warmup_steps": 203,
"weight_decay": 2.8436289860950645e-08
}
}