| { | |
| "best_metric": 0.42603132128715515, | |
| "best_model_checkpoint": "./mistral/20-04-24-Weni-WeniGPT-Agents-Mistral-1.0.6-SFT-1.0.5-DPO_Experiment on DPO with other hyperparameters and best SFT model of WeniGPT-2_max_steps-366_batch_4_2024-04-20_ppid_9/checkpoint-360", | |
| "epoch": 5.853658536585366, | |
| "eval_steps": 30, | |
| "global_step": 360, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.16, | |
| "grad_norm": 8.378021240234375, | |
| "learning_rate": 4.0909090909090915e-06, | |
| "logits/chosen": -1.830958604812622, | |
| "logits/rejected": -1.8507845401763916, | |
| "logps/chosen": -28.701984405517578, | |
| "logps/rejected": -54.28569793701172, | |
| "loss": 0.6924, | |
| "rewards/accuracies": 0.20000000298023224, | |
| "rewards/chosen": 0.0008967495523393154, | |
| "rewards/margins": 0.0014666033675894141, | |
| "rewards/rejected": -0.0005698538152500987, | |
| "step": 10 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "grad_norm": 5.193418502807617, | |
| "learning_rate": 4.887323943661972e-06, | |
| "logits/chosen": -1.7550897598266602, | |
| "logits/rejected": -1.770708680152893, | |
| "logps/chosen": -47.344207763671875, | |
| "logps/rejected": -64.0368423461914, | |
| "loss": 0.6852, | |
| "rewards/accuracies": 0.4000000059604645, | |
| "rewards/chosen": 0.017231885343790054, | |
| "rewards/margins": 0.01606021076440811, | |
| "rewards/rejected": 0.0011716745793819427, | |
| "step": 20 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "grad_norm": 7.308932304382324, | |
| "learning_rate": 4.746478873239437e-06, | |
| "logits/chosen": -1.781267762184143, | |
| "logits/rejected": -1.8114898204803467, | |
| "logps/chosen": -54.274559020996094, | |
| "logps/rejected": -95.20500183105469, | |
| "loss": 0.6635, | |
| "rewards/accuracies": 0.5, | |
| "rewards/chosen": 0.0641159638762474, | |
| "rewards/margins": 0.061691801995038986, | |
| "rewards/rejected": 0.0024241588544100523, | |
| "step": 30 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "eval_logits/chosen": -1.7831767797470093, | |
| "eval_logits/rejected": -1.8043663501739502, | |
| "eval_logps/chosen": -55.16960906982422, | |
| "eval_logps/rejected": -97.32585144042969, | |
| "eval_loss": 0.6523757576942444, | |
| "eval_rewards/accuracies": 0.4642857015132904, | |
| "eval_rewards/chosen": 0.09036973863840103, | |
| "eval_rewards/margins": 0.08673857897520065, | |
| "eval_rewards/rejected": 0.0036311547737568617, | |
| "eval_runtime": 8.141, | |
| "eval_samples_per_second": 3.439, | |
| "eval_steps_per_second": 1.72, | |
| "step": 30 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "grad_norm": 0.0, | |
| "learning_rate": 4.6056338028169015e-06, | |
| "logits/chosen": -1.889905333518982, | |
| "logits/rejected": -1.9024461507797241, | |
| "logps/chosen": -27.918941497802734, | |
| "logps/rejected": -42.093284606933594, | |
| "loss": 0.668, | |
| "rewards/accuracies": 0.25, | |
| "rewards/chosen": 0.054457180202007294, | |
| "rewards/margins": 0.0539846234023571, | |
| "rewards/rejected": 0.0004725646285805851, | |
| "step": 40 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "grad_norm": 8.53225326538086, | |
| "learning_rate": 4.464788732394367e-06, | |
| "logits/chosen": -1.8278567790985107, | |
| "logits/rejected": -1.849957823753357, | |
| "logps/chosen": -43.8238639831543, | |
| "logps/rejected": -68.02179718017578, | |
| "loss": 0.6358, | |
| "rewards/accuracies": 0.3499999940395355, | |
| "rewards/chosen": 0.13941256701946259, | |
| "rewards/margins": 0.13133978843688965, | |
| "rewards/rejected": 0.008072790689766407, | |
| "step": 50 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "grad_norm": 9.436968803405762, | |
| "learning_rate": 4.3239436619718315e-06, | |
| "logits/chosen": -1.805991768836975, | |
| "logits/rejected": -1.8437427282333374, | |
| "logps/chosen": -43.8873291015625, | |
| "logps/rejected": -95.2943115234375, | |
| "loss": 0.6026, | |
| "rewards/accuracies": 0.44999998807907104, | |
| "rewards/chosen": 0.18793432414531708, | |
| "rewards/margins": 0.21308371424674988, | |
| "rewards/rejected": -0.025149401277303696, | |
| "step": 60 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "eval_logits/chosen": -1.7877694368362427, | |
| "eval_logits/rejected": -1.8098936080932617, | |
| "eval_logps/chosen": -53.567203521728516, | |
| "eval_logps/rejected": -97.33795928955078, | |
| "eval_loss": 0.5890871286392212, | |
| "eval_rewards/accuracies": 0.4642857015132904, | |
| "eval_rewards/chosen": 0.25061002373695374, | |
| "eval_rewards/margins": 0.2481890469789505, | |
| "eval_rewards/rejected": 0.002420984674245119, | |
| "eval_runtime": 8.1404, | |
| "eval_samples_per_second": 3.44, | |
| "eval_steps_per_second": 1.72, | |
| "step": 60 | |
| }, | |
| { | |
| "epoch": 1.14, | |
| "grad_norm": 0.0, | |
| "learning_rate": 4.183098591549296e-06, | |
| "logits/chosen": -1.8344879150390625, | |
| "logits/rejected": -1.8489716053009033, | |
| "logps/chosen": -40.38930892944336, | |
| "logps/rejected": -60.9084358215332, | |
| "loss": 0.6031, | |
| "rewards/accuracies": 0.375, | |
| "rewards/chosen": 0.19739331305027008, | |
| "rewards/margins": 0.22638121247291565, | |
| "rewards/rejected": -0.028987903147935867, | |
| "step": 70 | |
| }, | |
| { | |
| "epoch": 1.3, | |
| "grad_norm": 5.49536657333374, | |
| "learning_rate": 4.042253521126761e-06, | |
| "logits/chosen": -1.7903095483779907, | |
| "logits/rejected": -1.8362411260604858, | |
| "logps/chosen": -44.288116455078125, | |
| "logps/rejected": -90.21073913574219, | |
| "loss": 0.5357, | |
| "rewards/accuracies": 0.4749999940395355, | |
| "rewards/chosen": 0.34061312675476074, | |
| "rewards/margins": 0.40679749846458435, | |
| "rewards/rejected": -0.06618441641330719, | |
| "step": 80 | |
| }, | |
| { | |
| "epoch": 1.46, | |
| "grad_norm": 13.401692390441895, | |
| "learning_rate": 3.901408450704225e-06, | |
| "logits/chosen": -1.8004281520843506, | |
| "logits/rejected": -1.8247934579849243, | |
| "logps/chosen": -42.32465362548828, | |
| "logps/rejected": -70.9749984741211, | |
| "loss": 0.5387, | |
| "rewards/accuracies": 0.4749999940395355, | |
| "rewards/chosen": 0.3678433299064636, | |
| "rewards/margins": 0.4186524450778961, | |
| "rewards/rejected": -0.05080908536911011, | |
| "step": 90 | |
| }, | |
| { | |
| "epoch": 1.46, | |
| "eval_logits/chosen": -1.7943389415740967, | |
| "eval_logits/rejected": -1.8181126117706299, | |
| "eval_logps/chosen": -51.677486419677734, | |
| "eval_logps/rejected": -97.63689422607422, | |
| "eval_loss": 0.529485821723938, | |
| "eval_rewards/accuracies": 0.4642857015132904, | |
| "eval_rewards/chosen": 0.4395819306373596, | |
| "eval_rewards/margins": 0.4670555889606476, | |
| "eval_rewards/rejected": -0.027473628520965576, | |
| "eval_runtime": 8.1412, | |
| "eval_samples_per_second": 3.439, | |
| "eval_steps_per_second": 1.72, | |
| "step": 90 | |
| }, | |
| { | |
| "epoch": 1.63, | |
| "grad_norm": 5.040858745574951, | |
| "learning_rate": 3.7605633802816903e-06, | |
| "logits/chosen": -1.8601042032241821, | |
| "logits/rejected": -1.8790462017059326, | |
| "logps/chosen": -43.77570343017578, | |
| "logps/rejected": -70.64997863769531, | |
| "loss": 0.5466, | |
| "rewards/accuracies": 0.4000000059604645, | |
| "rewards/chosen": 0.36673134565353394, | |
| "rewards/margins": 0.42903366684913635, | |
| "rewards/rejected": -0.06230226159095764, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 1.79, | |
| "grad_norm": 11.182683944702148, | |
| "learning_rate": 3.6197183098591553e-06, | |
| "logits/chosen": -1.8602203130722046, | |
| "logits/rejected": -1.8786903619766235, | |
| "logps/chosen": -29.601736068725586, | |
| "logps/rejected": -66.1338882446289, | |
| "loss": 0.6003, | |
| "rewards/accuracies": 0.2750000059604645, | |
| "rewards/chosen": 0.3122637867927551, | |
| "rewards/margins": 0.2756831645965576, | |
| "rewards/rejected": 0.03658062964677811, | |
| "step": 110 | |
| }, | |
| { | |
| "epoch": 1.95, | |
| "grad_norm": 3.9169583320617676, | |
| "learning_rate": 3.47887323943662e-06, | |
| "logits/chosen": -1.8304624557495117, | |
| "logits/rejected": -1.8451646566390991, | |
| "logps/chosen": -31.413599014282227, | |
| "logps/rejected": -56.841880798339844, | |
| "loss": 0.6033, | |
| "rewards/accuracies": 0.2750000059604645, | |
| "rewards/chosen": 0.21899382770061493, | |
| "rewards/margins": 0.2744571566581726, | |
| "rewards/rejected": -0.05546332150697708, | |
| "step": 120 | |
| }, | |
| { | |
| "epoch": 1.95, | |
| "eval_logits/chosen": -1.80086350440979, | |
| "eval_logits/rejected": -1.8260576725006104, | |
| "eval_logps/chosen": -50.32191848754883, | |
| "eval_logps/rejected": -98.02101135253906, | |
| "eval_loss": 0.49604225158691406, | |
| "eval_rewards/accuracies": 0.4642857015132904, | |
| "eval_rewards/chosen": 0.5751391053199768, | |
| "eval_rewards/margins": 0.6410244107246399, | |
| "eval_rewards/rejected": -0.0658852607011795, | |
| "eval_runtime": 8.1445, | |
| "eval_samples_per_second": 3.438, | |
| "eval_steps_per_second": 1.719, | |
| "step": 120 | |
| }, | |
| { | |
| "epoch": 2.11, | |
| "grad_norm": 1.4047716856002808, | |
| "learning_rate": 3.338028169014085e-06, | |
| "logits/chosen": -1.8776130676269531, | |
| "logits/rejected": -1.8995519876480103, | |
| "logps/chosen": -22.69371795654297, | |
| "logps/rejected": -53.5282096862793, | |
| "loss": 0.5611, | |
| "rewards/accuracies": 0.25, | |
| "rewards/chosen": 0.35938918590545654, | |
| "rewards/margins": 0.5045264959335327, | |
| "rewards/rejected": -0.14513733983039856, | |
| "step": 130 | |
| }, | |
| { | |
| "epoch": 2.28, | |
| "grad_norm": 0.7528722882270813, | |
| "learning_rate": 3.1971830985915496e-06, | |
| "logits/chosen": -1.8126357793807983, | |
| "logits/rejected": -1.832371711730957, | |
| "logps/chosen": -38.33379364013672, | |
| "logps/rejected": -67.96979522705078, | |
| "loss": 0.5142, | |
| "rewards/accuracies": 0.375, | |
| "rewards/chosen": 0.5568062663078308, | |
| "rewards/margins": 0.6818712949752808, | |
| "rewards/rejected": -0.12506499886512756, | |
| "step": 140 | |
| }, | |
| { | |
| "epoch": 2.44, | |
| "grad_norm": 3.405579090118408, | |
| "learning_rate": 3.056338028169014e-06, | |
| "logits/chosen": -1.8196109533309937, | |
| "logits/rejected": -1.8556429147720337, | |
| "logps/chosen": -36.78864669799805, | |
| "logps/rejected": -83.05890655517578, | |
| "loss": 0.5042, | |
| "rewards/accuracies": 0.4000000059604645, | |
| "rewards/chosen": 0.542107105255127, | |
| "rewards/margins": 0.6411095857620239, | |
| "rewards/rejected": -0.09900249540805817, | |
| "step": 150 | |
| }, | |
| { | |
| "epoch": 2.44, | |
| "eval_logits/chosen": -1.805869698524475, | |
| "eval_logits/rejected": -1.8330577611923218, | |
| "eval_logps/chosen": -49.10601043701172, | |
| "eval_logps/rejected": -98.84068298339844, | |
| "eval_loss": 0.4709201455116272, | |
| "eval_rewards/accuracies": 0.4642857015132904, | |
| "eval_rewards/chosen": 0.6967297196388245, | |
| "eval_rewards/margins": 0.8445812463760376, | |
| "eval_rewards/rejected": -0.1478516012430191, | |
| "eval_runtime": 8.1382, | |
| "eval_samples_per_second": 3.441, | |
| "eval_steps_per_second": 1.72, | |
| "step": 150 | |
| }, | |
| { | |
| "epoch": 2.6, | |
| "grad_norm": 7.778740882873535, | |
| "learning_rate": 2.915492957746479e-06, | |
| "logits/chosen": -1.848589301109314, | |
| "logits/rejected": -1.8790754079818726, | |
| "logps/chosen": -36.49171447753906, | |
| "logps/rejected": -72.55968475341797, | |
| "loss": 0.4927, | |
| "rewards/accuracies": 0.44999998807907104, | |
| "rewards/chosen": 0.49555450677871704, | |
| "rewards/margins": 0.6891830563545227, | |
| "rewards/rejected": -0.1936284601688385, | |
| "step": 160 | |
| }, | |
| { | |
| "epoch": 2.76, | |
| "grad_norm": 4.058627605438232, | |
| "learning_rate": 2.774647887323944e-06, | |
| "logits/chosen": -1.812421441078186, | |
| "logits/rejected": -1.8415311574935913, | |
| "logps/chosen": -45.62999725341797, | |
| "logps/rejected": -87.85527038574219, | |
| "loss": 0.4541, | |
| "rewards/accuracies": 0.4749999940395355, | |
| "rewards/chosen": 0.7084562182426453, | |
| "rewards/margins": 0.9553689956665039, | |
| "rewards/rejected": -0.24691279232501984, | |
| "step": 170 | |
| }, | |
| { | |
| "epoch": 2.93, | |
| "grad_norm": 0.0, | |
| "learning_rate": 2.6338028169014084e-06, | |
| "logits/chosen": -1.8475942611694336, | |
| "logits/rejected": -1.8678725957870483, | |
| "logps/chosen": -40.53328323364258, | |
| "logps/rejected": -64.86616516113281, | |
| "loss": 0.5087, | |
| "rewards/accuracies": 0.375, | |
| "rewards/chosen": 0.5022943019866943, | |
| "rewards/margins": 0.7252141833305359, | |
| "rewards/rejected": -0.22291991114616394, | |
| "step": 180 | |
| }, | |
| { | |
| "epoch": 2.93, | |
| "eval_logits/chosen": -1.8136398792266846, | |
| "eval_logits/rejected": -1.8424787521362305, | |
| "eval_logps/chosen": -48.19547653198242, | |
| "eval_logps/rejected": -99.7900161743164, | |
| "eval_loss": 0.4541548192501068, | |
| "eval_rewards/accuracies": 0.4642857015132904, | |
| "eval_rewards/chosen": 0.7877826690673828, | |
| "eval_rewards/margins": 1.0305674076080322, | |
| "eval_rewards/rejected": -0.24278469383716583, | |
| "eval_runtime": 8.1397, | |
| "eval_samples_per_second": 3.44, | |
| "eval_steps_per_second": 1.72, | |
| "step": 180 | |
| }, | |
| { | |
| "epoch": 3.09, | |
| "grad_norm": 2.1618106365203857, | |
| "learning_rate": 2.4929577464788734e-06, | |
| "logits/chosen": -1.876151442527771, | |
| "logits/rejected": -1.9132931232452393, | |
| "logps/chosen": -38.02617645263672, | |
| "logps/rejected": -84.4028549194336, | |
| "loss": 0.4372, | |
| "rewards/accuracies": 0.44999998807907104, | |
| "rewards/chosen": 0.723468005657196, | |
| "rewards/margins": 1.1590527296066284, | |
| "rewards/rejected": -0.43558478355407715, | |
| "step": 190 | |
| }, | |
| { | |
| "epoch": 3.25, | |
| "grad_norm": 0.0, | |
| "learning_rate": 2.352112676056338e-06, | |
| "logits/chosen": -1.8977773189544678, | |
| "logits/rejected": -1.9120800495147705, | |
| "logps/chosen": -38.097923278808594, | |
| "logps/rejected": -55.17757034301758, | |
| "loss": 0.4778, | |
| "rewards/accuracies": 0.375, | |
| "rewards/chosen": 0.5927585959434509, | |
| "rewards/margins": 0.8746024370193481, | |
| "rewards/rejected": -0.2818438410758972, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 3.41, | |
| "grad_norm": 7.095726013183594, | |
| "learning_rate": 2.211267605633803e-06, | |
| "logits/chosen": -1.8508259057998657, | |
| "logits/rejected": -1.8876402378082275, | |
| "logps/chosen": -33.23273468017578, | |
| "logps/rejected": -79.0272445678711, | |
| "loss": 0.4874, | |
| "rewards/accuracies": 0.375, | |
| "rewards/chosen": 0.5798195600509644, | |
| "rewards/margins": 0.9200228452682495, | |
| "rewards/rejected": -0.34020328521728516, | |
| "step": 210 | |
| }, | |
| { | |
| "epoch": 3.41, | |
| "eval_logits/chosen": -1.821912407875061, | |
| "eval_logits/rejected": -1.8520457744598389, | |
| "eval_logps/chosen": -47.6314697265625, | |
| "eval_logps/rejected": -100.92195129394531, | |
| "eval_loss": 0.4427572786808014, | |
| "eval_rewards/accuracies": 0.4642857015132904, | |
| "eval_rewards/chosen": 0.8441829681396484, | |
| "eval_rewards/margins": 1.2001608610153198, | |
| "eval_rewards/rejected": -0.35597795248031616, | |
| "eval_runtime": 8.1451, | |
| "eval_samples_per_second": 3.438, | |
| "eval_steps_per_second": 1.719, | |
| "step": 210 | |
| }, | |
| { | |
| "epoch": 3.58, | |
| "grad_norm": 4.825575351715088, | |
| "learning_rate": 2.0704225352112676e-06, | |
| "logits/chosen": -1.889478325843811, | |
| "logits/rejected": -1.9106714725494385, | |
| "logps/chosen": -30.769512176513672, | |
| "logps/rejected": -68.92756652832031, | |
| "loss": 0.5277, | |
| "rewards/accuracies": 0.32499998807907104, | |
| "rewards/chosen": 0.379320353269577, | |
| "rewards/margins": 0.6602964401245117, | |
| "rewards/rejected": -0.28097596764564514, | |
| "step": 220 | |
| }, | |
| { | |
| "epoch": 3.74, | |
| "grad_norm": 5.236915588378906, | |
| "learning_rate": 1.9295774647887326e-06, | |
| "logits/chosen": -1.8926284313201904, | |
| "logits/rejected": -1.9087079763412476, | |
| "logps/chosen": -36.48774719238281, | |
| "logps/rejected": -59.29833221435547, | |
| "loss": 0.5176, | |
| "rewards/accuracies": 0.32499998807907104, | |
| "rewards/chosen": 0.6325365304946899, | |
| "rewards/margins": 0.867927074432373, | |
| "rewards/rejected": -0.2353905737400055, | |
| "step": 230 | |
| }, | |
| { | |
| "epoch": 3.9, | |
| "grad_norm": 1.3737443685531616, | |
| "learning_rate": 1.7887323943661974e-06, | |
| "logits/chosen": -1.7782018184661865, | |
| "logits/rejected": -1.8105701208114624, | |
| "logps/chosen": -41.42538833618164, | |
| "logps/rejected": -93.73129272460938, | |
| "loss": 0.4229, | |
| "rewards/accuracies": 0.4749999940395355, | |
| "rewards/chosen": 0.8450711369514465, | |
| "rewards/margins": 1.3813583850860596, | |
| "rewards/rejected": -0.5362871885299683, | |
| "step": 240 | |
| }, | |
| { | |
| "epoch": 3.9, | |
| "eval_logits/chosen": -1.826602816581726, | |
| "eval_logits/rejected": -1.8575078248977661, | |
| "eval_logps/chosen": -47.322914123535156, | |
| "eval_logps/rejected": -101.7520980834961, | |
| "eval_loss": 0.4358247220516205, | |
| "eval_rewards/accuracies": 0.4642857015132904, | |
| "eval_rewards/chosen": 0.8750395178794861, | |
| "eval_rewards/margins": 1.3140336275100708, | |
| "eval_rewards/rejected": -0.4389941692352295, | |
| "eval_runtime": 8.1403, | |
| "eval_samples_per_second": 3.44, | |
| "eval_steps_per_second": 1.72, | |
| "step": 240 | |
| }, | |
| { | |
| "epoch": 4.07, | |
| "grad_norm": 1.977386713027954, | |
| "learning_rate": 1.647887323943662e-06, | |
| "logits/chosen": -1.875792145729065, | |
| "logits/rejected": -1.8937476873397827, | |
| "logps/chosen": -25.06104278564453, | |
| "logps/rejected": -46.700584411621094, | |
| "loss": 0.5274, | |
| "rewards/accuracies": 0.30000001192092896, | |
| "rewards/chosen": 0.42553478479385376, | |
| "rewards/margins": 0.7891250252723694, | |
| "rewards/rejected": -0.3635903000831604, | |
| "step": 250 | |
| }, | |
| { | |
| "epoch": 4.23, | |
| "grad_norm": 3.320791244506836, | |
| "learning_rate": 1.5070422535211269e-06, | |
| "logits/chosen": -1.7908179759979248, | |
| "logits/rejected": -1.8309694528579712, | |
| "logps/chosen": -54.056663513183594, | |
| "logps/rejected": -108.03240966796875, | |
| "loss": 0.3569, | |
| "rewards/accuracies": 0.574999988079071, | |
| "rewards/chosen": 1.07839035987854, | |
| "rewards/margins": 1.643531084060669, | |
| "rewards/rejected": -0.5651407837867737, | |
| "step": 260 | |
| }, | |
| { | |
| "epoch": 4.39, | |
| "grad_norm": 4.999856948852539, | |
| "learning_rate": 1.3661971830985919e-06, | |
| "logits/chosen": -1.9177863597869873, | |
| "logits/rejected": -1.9476194381713867, | |
| "logps/chosen": -22.58294105529785, | |
| "logps/rejected": -61.99756622314453, | |
| "loss": 0.5295, | |
| "rewards/accuracies": 0.2750000059604645, | |
| "rewards/chosen": 0.5058903694152832, | |
| "rewards/margins": 0.8186748623847961, | |
| "rewards/rejected": -0.31278449296951294, | |
| "step": 270 | |
| }, | |
| { | |
| "epoch": 4.39, | |
| "eval_logits/chosen": -1.8289211988449097, | |
| "eval_logits/rejected": -1.860676646232605, | |
| "eval_logps/chosen": -47.04714584350586, | |
| "eval_logps/rejected": -102.3218994140625, | |
| "eval_loss": 0.43130752444267273, | |
| "eval_rewards/accuracies": 0.4642857015132904, | |
| "eval_rewards/chosen": 0.9026166200637817, | |
| "eval_rewards/margins": 1.3985893726348877, | |
| "eval_rewards/rejected": -0.4959728717803955, | |
| "eval_runtime": 8.1397, | |
| "eval_samples_per_second": 3.44, | |
| "eval_steps_per_second": 1.72, | |
| "step": 270 | |
| }, | |
| { | |
| "epoch": 4.55, | |
| "grad_norm": 0.36910170316696167, | |
| "learning_rate": 1.2253521126760565e-06, | |
| "logits/chosen": -1.881696105003357, | |
| "logits/rejected": -1.9073266983032227, | |
| "logps/chosen": -33.28648376464844, | |
| "logps/rejected": -68.246337890625, | |
| "loss": 0.4962, | |
| "rewards/accuracies": 0.32499998807907104, | |
| "rewards/chosen": 0.5577932000160217, | |
| "rewards/margins": 1.0508588552474976, | |
| "rewards/rejected": -0.4930656850337982, | |
| "step": 280 | |
| }, | |
| { | |
| "epoch": 4.72, | |
| "grad_norm": 11.332355499267578, | |
| "learning_rate": 1.084507042253521e-06, | |
| "logits/chosen": -1.8866857290267944, | |
| "logits/rejected": -1.900857925415039, | |
| "logps/chosen": -40.804874420166016, | |
| "logps/rejected": -71.67508697509766, | |
| "loss": 0.4851, | |
| "rewards/accuracies": 0.4000000059604645, | |
| "rewards/chosen": 0.7005030512809753, | |
| "rewards/margins": 1.2071340084075928, | |
| "rewards/rejected": -0.5066308379173279, | |
| "step": 290 | |
| }, | |
| { | |
| "epoch": 4.88, | |
| "grad_norm": 3.657494306564331, | |
| "learning_rate": 9.43661971830986e-07, | |
| "logits/chosen": -1.9023106098175049, | |
| "logits/rejected": -1.9253908395767212, | |
| "logps/chosen": -18.57657814025879, | |
| "logps/rejected": -53.88740158081055, | |
| "loss": 0.5466, | |
| "rewards/accuracies": 0.25, | |
| "rewards/chosen": 0.3226935565471649, | |
| "rewards/margins": 0.6567031145095825, | |
| "rewards/rejected": -0.33400958776474, | |
| "step": 300 | |
| }, | |
| { | |
| "epoch": 4.88, | |
| "eval_logits/chosen": -1.8308794498443604, | |
| "eval_logits/rejected": -1.8629435300827026, | |
| "eval_logps/chosen": -46.95443344116211, | |
| "eval_logps/rejected": -102.74605560302734, | |
| "eval_loss": 0.4291366934776306, | |
| "eval_rewards/accuracies": 0.4642857015132904, | |
| "eval_rewards/chosen": 0.9118875861167908, | |
| "eval_rewards/margins": 1.4502772092819214, | |
| "eval_rewards/rejected": -0.5383896827697754, | |
| "eval_runtime": 8.1441, | |
| "eval_samples_per_second": 3.438, | |
| "eval_steps_per_second": 1.719, | |
| "step": 300 | |
| }, | |
| { | |
| "epoch": 5.04, | |
| "grad_norm": 4.444954872131348, | |
| "learning_rate": 8.028169014084508e-07, | |
| "logits/chosen": -1.835021734237671, | |
| "logits/rejected": -1.858599066734314, | |
| "logps/chosen": -42.14970016479492, | |
| "logps/rejected": -86.938720703125, | |
| "loss": 0.4128, | |
| "rewards/accuracies": 0.4749999940395355, | |
| "rewards/chosen": 0.8941621780395508, | |
| "rewards/margins": 1.572546362876892, | |
| "rewards/rejected": -0.6783844232559204, | |
| "step": 310 | |
| }, | |
| { | |
| "epoch": 5.2, | |
| "grad_norm": 0.4182775616645813, | |
| "learning_rate": 6.619718309859155e-07, | |
| "logits/chosen": -1.8859401941299438, | |
| "logits/rejected": -1.910548448562622, | |
| "logps/chosen": -34.28424835205078, | |
| "logps/rejected": -77.3191146850586, | |
| "loss": 0.4465, | |
| "rewards/accuracies": 0.4000000059604645, | |
| "rewards/chosen": 0.739007294178009, | |
| "rewards/margins": 1.3678598403930664, | |
| "rewards/rejected": -0.6288524866104126, | |
| "step": 320 | |
| }, | |
| { | |
| "epoch": 5.37, | |
| "grad_norm": 2.8709957859973656e-06, | |
| "learning_rate": 5.211267605633803e-07, | |
| "logits/chosen": -1.7752397060394287, | |
| "logits/rejected": -1.8195409774780273, | |
| "logps/chosen": -42.48664855957031, | |
| "logps/rejected": -97.59371185302734, | |
| "loss": 0.4339, | |
| "rewards/accuracies": 0.5, | |
| "rewards/chosen": 0.8835798501968384, | |
| "rewards/margins": 1.3967663049697876, | |
| "rewards/rejected": -0.513186514377594, | |
| "step": 330 | |
| }, | |
| { | |
| "epoch": 5.37, | |
| "eval_logits/chosen": -1.8319826126098633, | |
| "eval_logits/rejected": -1.864353895187378, | |
| "eval_logps/chosen": -46.921607971191406, | |
| "eval_logps/rejected": -103.26231384277344, | |
| "eval_loss": 0.42683711647987366, | |
| "eval_rewards/accuracies": 0.4642857015132904, | |
| "eval_rewards/chosen": 0.9151698350906372, | |
| "eval_rewards/margins": 1.505185842514038, | |
| "eval_rewards/rejected": -0.5900159478187561, | |
| "eval_runtime": 8.1406, | |
| "eval_samples_per_second": 3.44, | |
| "eval_steps_per_second": 1.72, | |
| "step": 330 | |
| }, | |
| { | |
| "epoch": 5.53, | |
| "grad_norm": 0.23427560925483704, | |
| "learning_rate": 3.8028169014084507e-07, | |
| "logits/chosen": -1.8918192386627197, | |
| "logits/rejected": -1.9169318675994873, | |
| "logps/chosen": -35.16785430908203, | |
| "logps/rejected": -71.60049438476562, | |
| "loss": 0.4617, | |
| "rewards/accuracies": 0.375, | |
| "rewards/chosen": 0.5324742197990417, | |
| "rewards/margins": 1.1221383810043335, | |
| "rewards/rejected": -0.5896641612052917, | |
| "step": 340 | |
| }, | |
| { | |
| "epoch": 5.69, | |
| "grad_norm": 0.0, | |
| "learning_rate": 2.394366197183099e-07, | |
| "logits/chosen": -1.857642412185669, | |
| "logits/rejected": -1.888279676437378, | |
| "logps/chosen": -37.31398010253906, | |
| "logps/rejected": -90.64387512207031, | |
| "loss": 0.4569, | |
| "rewards/accuracies": 0.4000000059604645, | |
| "rewards/chosen": 0.7042752504348755, | |
| "rewards/margins": 1.4055907726287842, | |
| "rewards/rejected": -0.7013154625892639, | |
| "step": 350 | |
| }, | |
| { | |
| "epoch": 5.85, | |
| "grad_norm": 11.415884017944336, | |
| "learning_rate": 9.859154929577466e-08, | |
| "logits/chosen": -1.9216959476470947, | |
| "logits/rejected": -1.9308369159698486, | |
| "logps/chosen": -25.689884185791016, | |
| "logps/rejected": -36.790706634521484, | |
| "loss": 0.5438, | |
| "rewards/accuracies": 0.25, | |
| "rewards/chosen": 0.4890199303627014, | |
| "rewards/margins": 0.789040207862854, | |
| "rewards/rejected": -0.3000202775001526, | |
| "step": 360 | |
| }, | |
| { | |
| "epoch": 5.85, | |
| "eval_logits/chosen": -1.832722544670105, | |
| "eval_logits/rejected": -1.8652076721191406, | |
| "eval_logps/chosen": -46.90084457397461, | |
| "eval_logps/rejected": -103.44039154052734, | |
| "eval_loss": 0.42603132128715515, | |
| "eval_rewards/accuracies": 0.4642857015132904, | |
| "eval_rewards/chosen": 0.9172464609146118, | |
| "eval_rewards/margins": 1.5250685214996338, | |
| "eval_rewards/rejected": -0.6078222990036011, | |
| "eval_runtime": 8.1439, | |
| "eval_samples_per_second": 3.438, | |
| "eval_steps_per_second": 1.719, | |
| "step": 360 | |
| } | |
| ], | |
| "logging_steps": 10, | |
| "max_steps": 366, | |
| "num_input_tokens_seen": 0, | |
| "num_train_epochs": 6, | |
| "save_steps": 90, | |
| "total_flos": 0.0, | |
| "train_batch_size": 2, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |