| { | |
| "best_metric": 0.1542571794425746, | |
| "best_model_checkpoint": "esm2_t12_35M_qlora_ptm_sites_2023-10-12_18-42-03/checkpoint-18344", | |
| "epoch": 0.9999727438741858, | |
| "eval_steps": 500, | |
| "global_step": 18344, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.00037004824930228345, | |
| "loss": 0.3438, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.00036972270781686296, | |
| "loss": 0.1099, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.00036918056301062725, | |
| "loss": 0.0758, | |
| "step": 600 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.0003684224508650944, | |
| "loss": 0.0587, | |
| "step": 800 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.00036744926070966046, | |
| "loss": 0.05, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.0003662621341783412, | |
| "loss": 0.0433, | |
| "step": 1200 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.00036486246387053945, | |
| "loss": 0.0399, | |
| "step": 1400 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.0003632518917174086, | |
| "loss": 0.0373, | |
| "step": 1600 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.00036143230705572906, | |
| "loss": 0.0345, | |
| "step": 1800 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.0003594058444115571, | |
| "loss": 0.0361, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.000357174880996246, | |
| "loss": 0.0324, | |
| "step": 2200 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.000354742033917777, | |
| "loss": 0.0294, | |
| "step": 2400 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.00035211015711067114, | |
| "loss": 0.0307, | |
| "step": 2600 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.0003492823379880838, | |
| "loss": 0.0286, | |
| "step": 2800 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.00034626189382000884, | |
| "loss": 0.0258, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.0003430523678418417, | |
| "loss": 0.0286, | |
| "step": 3200 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.0003396575250978659, | |
| "loss": 0.0259, | |
| "step": 3400 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.0003360813480245388, | |
| "loss": 0.0254, | |
| "step": 3600 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.00033232803177875834, | |
| "loss": 0.0258, | |
| "step": 3800 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.00032840197931659074, | |
| "loss": 0.0242, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.00032432867750442764, | |
| "loss": 0.0232, | |
| "step": 4200 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.0003200936526736335, | |
| "loss": 0.0268, | |
| "step": 4400 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.0003156793663915513, | |
| "loss": 0.0254, | |
| "step": 4600 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.0003111118741797566, | |
| "loss": 0.0261, | |
| "step": 4800 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.0003063965340912947, | |
| "loss": 0.0258, | |
| "step": 5000 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.00030153887761722234, | |
| "loss": 0.0266, | |
| "step": 5200 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.00029654460319770276, | |
| "loss": 0.0228, | |
| "step": 5400 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.0002914195695372552, | |
| "loss": 0.0222, | |
| "step": 5600 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.00028616978873200133, | |
| "loss": 0.0217, | |
| "step": 5800 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.0002808014192169701, | |
| "loss": 0.0218, | |
| "step": 6000 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.0002753207585417347, | |
| "loss": 0.0218, | |
| "step": 6200 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.0002697342359828566, | |
| "loss": 0.0201, | |
| "step": 6400 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.00026404840500180257, | |
| "loss": 0.0237, | |
| "step": 6600 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.00025826993555718255, | |
| "loss": 0.0209, | |
| "step": 6800 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.0002524351301631664, | |
| "loss": 0.0227, | |
| "step": 7000 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.0002464921980395868, | |
| "loss": 0.0209, | |
| "step": 7200 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.00024047722236181782, | |
| "loss": 0.0177, | |
| "step": 7400 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.0002343972592029196, | |
| "loss": 0.0227, | |
| "step": 7600 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.00022825944087174108, | |
| "loss": 0.019, | |
| "step": 7800 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.00022207096754612101, | |
| "loss": 0.0186, | |
| "step": 8000 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.00021583909882647275, | |
| "loss": 0.0202, | |
| "step": 8200 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.0002095711452196616, | |
| "loss": 0.0205, | |
| "step": 8400 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.00020327445956316495, | |
| "loss": 0.0193, | |
| "step": 8600 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 0.00019695642839957393, | |
| "loss": 0.0199, | |
| "step": 8800 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 0.0001906561455150737, | |
| "loss": 0.0224, | |
| "step": 9000 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 0.00018431768846772684, | |
| "loss": 0.0199, | |
| "step": 9200 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 0.00017798012380234812, | |
| "loss": 0.0196, | |
| "step": 9400 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 0.00017165088601604907, | |
| "loss": 0.0219, | |
| "step": 9600 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 0.00016533739983781094, | |
| "loss": 0.0211, | |
| "step": 9800 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 0.000159047071518652, | |
| "loss": 0.0177, | |
| "step": 10000 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 0.00015278728014347137, | |
| "loss": 0.0182, | |
| "step": 10200 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 0.00014656536897476115, | |
| "loss": 0.021, | |
| "step": 10400 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 0.00014038863683834082, | |
| "loss": 0.018, | |
| "step": 10600 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 0.00013426432956121996, | |
| "loss": 0.0183, | |
| "step": 10800 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 0.00012819963147163302, | |
| "loss": 0.0196, | |
| "step": 11000 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 0.00012220165697121683, | |
| "loss": 0.0167, | |
| "step": 11200 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 0.00011627744218921885, | |
| "loss": 0.0169, | |
| "step": 11400 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 0.00011043393672852448, | |
| "loss": 0.0178, | |
| "step": 11600 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 0.00010467799551318781, | |
| "loss": 0.0188, | |
| "step": 11800 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 9.901637074702851e-05, | |
| "loss": 0.0195, | |
| "step": 12000 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 9.345570399272756e-05, | |
| "loss": 0.0214, | |
| "step": 12200 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 8.80025183807155e-05, | |
| "loss": 0.0177, | |
| "step": 12400 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 8.26632109569906e-05, | |
| "loss": 0.0189, | |
| "step": 12600 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 7.744404517884536e-05, | |
| "loss": 0.0197, | |
| "step": 12800 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 7.235114356730338e-05, | |
| "loss": 0.016, | |
| "step": 13000 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 6.739048052488678e-05, | |
| "loss": 0.0186, | |
| "step": 13200 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 6.256787532713881e-05, | |
| "loss": 0.0183, | |
| "step": 13400 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 5.7888985296123274e-05, | |
| "loss": 0.0171, | |
| "step": 13600 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 5.3359299163909725e-05, | |
| "loss": 0.0174, | |
| "step": 13800 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 4.898413063382821e-05, | |
| "loss": 0.0172, | |
| "step": 14000 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 4.476861214704764e-05, | |
| "loss": 0.0181, | |
| "step": 14200 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 4.071768886179024e-05, | |
| "loss": 0.0162, | |
| "step": 14400 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 3.683611285224441e-05, | |
| "loss": 0.0173, | |
| "step": 14600 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 3.312843753398155e-05, | |
| "loss": 0.0182, | |
| "step": 14800 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 2.9616209114343275e-05, | |
| "loss": 0.0184, | |
| "step": 15000 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 2.6268252423346314e-05, | |
| "loss": 0.0173, | |
| "step": 15200 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 2.310659341395623e-05, | |
| "loss": 0.0193, | |
| "step": 15400 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 2.0134940978470748e-05, | |
| "loss": 0.0156, | |
| "step": 15600 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.737018511568027e-05, | |
| "loss": 0.0193, | |
| "step": 15800 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 1.4787785356205422e-05, | |
| "loss": 0.0171, | |
| "step": 16000 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 1.2405150833109309e-05, | |
| "loss": 0.0184, | |
| "step": 16200 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 1.0225076577356552e-05, | |
| "loss": 0.0171, | |
| "step": 16400 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 8.25011999965819e-06, | |
| "loss": 0.0181, | |
| "step": 16600 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 6.482597890413681e-06, | |
| "loss": 0.0162, | |
| "step": 16800 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 4.924583701920537e-06, | |
| "loss": 0.0175, | |
| "step": 17000 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 3.5779051160400913e-06, | |
| "loss": 0.017, | |
| "step": 17200 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 2.4441419001728858e-06, | |
| "loss": 0.0196, | |
| "step": 17400 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 1.524624054058352e-06, | |
| "loss": 0.0184, | |
| "step": 17600 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 8.204302495731164e-07, | |
| "loss": 0.0176, | |
| "step": 17800 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 3.3238656535812453e-07, | |
| "loss": 0.018, | |
| "step": 18000 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 6.106551775873308e-08, | |
| "loss": 0.0154, | |
| "step": 18200 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_accuracy": 0.9762591331328516, | |
| "eval_auc": 0.8833701456278934, | |
| "eval_f1": 0.1542571794425746, | |
| "eval_loss": 0.28556737303733826, | |
| "eval_mcc": 0.25511446421928063, | |
| "eval_precision": 0.08547382057474782, | |
| "eval_recall": 0.7899691877651231, | |
| "eval_runtime": 10582.0535, | |
| "eval_samples_per_second": 42.891, | |
| "eval_steps_per_second": 0.894, | |
| "step": 18344 | |
| } | |
| ], | |
| "logging_steps": 200, | |
| "max_steps": 18344, | |
| "num_train_epochs": 1, | |
| "save_steps": 500, | |
| "total_flos": 3.5283848306530406e+17, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |