diff --git "a/checkpoint-910/trainer_state.json" "b/checkpoint-910/trainer_state.json" new file mode 100644--- /dev/null +++ "b/checkpoint-910/trainer_state.json" @@ -0,0 +1,13683 @@ +{ + "best_metric": null, + "best_model_checkpoint": null, + "epoch": 1.9978046103183313, + "eval_steps": 500, + "global_step": 910, + "is_hyper_param_search": false, + "is_local_process_zero": true, + "is_world_process_zero": true, + "log_history": [ + { + "epoch": 0.0021953896816684962, + "grad_norm": 1.1034454107284546, + "learning_rate": 5.0000000000000004e-08, + "logits/chosen": -2.2846126556396484, + "logits/rejected": -2.358996868133545, + "logps/chosen": -12.224668502807617, + "logps/rejected": -9.611772537231445, + "loss": 0.953, + "rewards/accuracies": 0.0, + "rewards/chosen": 0.0, + "rewards/margins": 0.0, + "rewards/rejected": 0.0, + "step": 1 + }, + { + "epoch": 0.0043907793633369925, + "grad_norm": 1.043821096420288, + "learning_rate": 1.0000000000000001e-07, + "logits/chosen": -2.293553352355957, + "logits/rejected": -2.2400941848754883, + "logps/chosen": -15.258804321289062, + "logps/rejected": -14.146197319030762, + "loss": 1.0789, + "rewards/accuracies": 0.0, + "rewards/chosen": 0.0, + "rewards/margins": 0.0, + "rewards/rejected": 0.0, + "step": 2 + }, + { + "epoch": 0.006586169045005488, + "grad_norm": 1.0436196327209473, + "learning_rate": 1.5000000000000002e-07, + "logits/chosen": -2.3567137718200684, + "logits/rejected": -2.3624215126037598, + "logps/chosen": -10.726156234741211, + "logps/rejected": -9.462532043457031, + "loss": 0.9529, + "rewards/accuracies": 0.40625, + "rewards/chosen": 0.002363908104598522, + "rewards/margins": -0.003392255399376154, + "rewards/rejected": 0.005756163503974676, + "step": 3 + }, + { + "epoch": 0.008781558726673985, + "grad_norm": 1.0475765466690063, + "learning_rate": 2.0000000000000002e-07, + "logits/chosen": -2.2790515422821045, + "logits/rejected": -2.2733278274536133, + "logps/chosen": -12.457335472106934, + "logps/rejected": -11.852964401245117, + "loss": 0.9939, + "rewards/accuracies": 0.53125, + "rewards/chosen": 0.0019623569678515196, + "rewards/margins": 0.005730569828301668, + "rewards/rejected": -0.003768211929127574, + "step": 4 + }, + { + "epoch": 0.010976948408342482, + "grad_norm": 1.0972965955734253, + "learning_rate": 2.5000000000000004e-07, + "logits/chosen": -2.3039631843566895, + "logits/rejected": -2.3130176067352295, + "logps/chosen": -12.775261878967285, + "logps/rejected": -12.0330171585083, + "loss": 1.0389, + "rewards/accuracies": 0.53125, + "rewards/chosen": -0.0011096245143562555, + "rewards/margins": -0.004510534927248955, + "rewards/rejected": 0.0034009101800620556, + "step": 5 + }, + { + "epoch": 0.013172338090010977, + "grad_norm": 0.9643481373786926, + "learning_rate": 3.0000000000000004e-07, + "logits/chosen": -2.324848175048828, + "logits/rejected": -2.429684638977051, + "logps/chosen": -18.116445541381836, + "logps/rejected": -15.265850067138672, + "loss": 1.0742, + "rewards/accuracies": 0.5, + "rewards/chosen": -0.0029701264575123787, + "rewards/margins": -0.003988922107964754, + "rewards/rejected": 0.0010187956504523754, + "step": 6 + }, + { + "epoch": 0.015367727771679473, + "grad_norm": 1.1616390943527222, + "learning_rate": 3.5000000000000004e-07, + "logits/chosen": -2.2589993476867676, + "logits/rejected": -2.3558740615844727, + "logps/chosen": -13.876771926879883, + "logps/rejected": -13.108857154846191, + "loss": 1.0832, + "rewards/accuracies": 0.5625, + "rewards/chosen": 0.0004574268823489547, + "rewards/margins": 0.00017213402315974236, + "rewards/rejected": 0.00028529250994324684, + "step": 7 + }, + { + "epoch": 0.01756311745334797, + "grad_norm": 1.283825397491455, + "learning_rate": 4.0000000000000003e-07, + "logits/chosen": -2.4278955459594727, + "logits/rejected": -2.3070013523101807, + "logps/chosen": -11.154717445373535, + "logps/rejected": -9.15652084350586, + "loss": 0.9298, + "rewards/accuracies": 0.4375, + "rewards/chosen": 0.00048030121251940727, + "rewards/margins": -0.006442471407353878, + "rewards/rejected": 0.006922773085534573, + "step": 8 + }, + { + "epoch": 0.019758507135016465, + "grad_norm": 1.4349981546401978, + "learning_rate": 4.5000000000000003e-07, + "logits/chosen": -2.3945982456207275, + "logits/rejected": -2.3926279544830322, + "logps/chosen": -12.6943359375, + "logps/rejected": -13.573047637939453, + "loss": 1.0278, + "rewards/accuracies": 0.53125, + "rewards/chosen": -3.980204928666353e-05, + "rewards/margins": 0.001762162777595222, + "rewards/rejected": -0.0018019648268818855, + "step": 9 + }, + { + "epoch": 0.021953896816684963, + "grad_norm": 1.1675068140029907, + "learning_rate": 5.000000000000001e-07, + "logits/chosen": -2.3849003314971924, + "logits/rejected": -2.360934257507324, + "logps/chosen": -18.73386001586914, + "logps/rejected": -16.180574417114258, + "loss": 1.1387, + "rewards/accuracies": 0.40625, + "rewards/chosen": -0.0015319950180128217, + "rewards/margins": -0.002129682805389166, + "rewards/rejected": 0.0005976876709610224, + "step": 10 + }, + { + "epoch": 0.024149286498353458, + "grad_norm": 1.2748152017593384, + "learning_rate": 5.5e-07, + "logits/chosen": -2.306387424468994, + "logits/rejected": -2.354555130004883, + "logps/chosen": -21.361164093017578, + "logps/rejected": -16.099632263183594, + "loss": 1.0941, + "rewards/accuracies": 0.5625, + "rewards/chosen": 0.00022546492982655764, + "rewards/margins": 0.004793249536305666, + "rewards/rejected": -0.004567783325910568, + "step": 11 + }, + { + "epoch": 0.026344676180021953, + "grad_norm": 1.2641266584396362, + "learning_rate": 6.000000000000001e-07, + "logits/chosen": -2.3452396392822266, + "logits/rejected": -2.358135223388672, + "logps/chosen": -11.315267562866211, + "logps/rejected": -11.734415054321289, + "loss": 1.0026, + "rewards/accuracies": 0.71875, + "rewards/chosen": 0.00597212091088295, + "rewards/margins": 0.010203271172940731, + "rewards/rejected": -0.004231149330735207, + "step": 12 + }, + { + "epoch": 0.02854006586169045, + "grad_norm": 0.9660422801971436, + "learning_rate": 6.5e-07, + "logits/chosen": -2.441119909286499, + "logits/rejected": -2.358609676361084, + "logps/chosen": -10.97406005859375, + "logps/rejected": -13.421890258789062, + "loss": 1.0251, + "rewards/accuracies": 0.4375, + "rewards/chosen": 0.0005808405112475157, + "rewards/margins": -0.00016933592269197106, + "rewards/rejected": 0.000750176259316504, + "step": 13 + }, + { + "epoch": 0.030735455543358946, + "grad_norm": 1.2691547870635986, + "learning_rate": 7.000000000000001e-07, + "logits/chosen": -2.3512799739837646, + "logits/rejected": -2.320150375366211, + "logps/chosen": -15.537447929382324, + "logps/rejected": -12.599323272705078, + "loss": 1.0558, + "rewards/accuracies": 0.34375, + "rewards/chosen": 0.0010906599927693605, + "rewards/margins": -0.0010261544957756996, + "rewards/rejected": 0.0021168149542063475, + "step": 14 + }, + { + "epoch": 0.03293084522502744, + "grad_norm": 1.1735515594482422, + "learning_rate": 7.5e-07, + "logits/chosen": -2.282367706298828, + "logits/rejected": -2.344204902648926, + "logps/chosen": -13.633934020996094, + "logps/rejected": -12.655315399169922, + "loss": 1.0283, + "rewards/accuracies": 0.59375, + "rewards/chosen": 0.0009840509155765176, + "rewards/margins": 0.0035789774265140295, + "rewards/rejected": -0.0025949266273528337, + "step": 15 + }, + { + "epoch": 0.03512623490669594, + "grad_norm": 1.2377164363861084, + "learning_rate": 8.000000000000001e-07, + "logits/chosen": -2.357888698577881, + "logits/rejected": -2.3939902782440186, + "logps/chosen": -16.876571655273438, + "logps/rejected": -16.07628631591797, + "loss": 1.1144, + "rewards/accuracies": 0.53125, + "rewards/chosen": 0.0009345448925159872, + "rewards/margins": 2.571078948676586e-05, + "rewards/rejected": 0.0009088342194445431, + "step": 16 + }, + { + "epoch": 0.03732162458836443, + "grad_norm": 1.1274445056915283, + "learning_rate": 8.500000000000001e-07, + "logits/chosen": -2.3163723945617676, + "logits/rejected": -2.316263198852539, + "logps/chosen": -19.3470401763916, + "logps/rejected": -18.514925003051758, + "loss": 1.1999, + "rewards/accuracies": 0.625, + "rewards/chosen": 0.003927948419004679, + "rewards/margins": 0.006356396712362766, + "rewards/rejected": -0.002428448526188731, + "step": 17 + }, + { + "epoch": 0.03951701427003293, + "grad_norm": 1.112360954284668, + "learning_rate": 9.000000000000001e-07, + "logits/chosen": -2.395773410797119, + "logits/rejected": -2.353379011154175, + "logps/chosen": -12.884025573730469, + "logps/rejected": -10.73438835144043, + "loss": 0.9752, + "rewards/accuracies": 0.375, + "rewards/chosen": 0.00025042565539479256, + "rewards/margins": 0.0016602440737187862, + "rewards/rejected": -0.0014098185347393155, + "step": 18 + }, + { + "epoch": 0.04171240395170143, + "grad_norm": 1.1286046504974365, + "learning_rate": 9.500000000000001e-07, + "logits/chosen": -2.387399673461914, + "logits/rejected": -2.3302669525146484, + "logps/chosen": -10.059947967529297, + "logps/rejected": -11.219734191894531, + "loss": 1.0013, + "rewards/accuracies": 0.40625, + "rewards/chosen": -0.0020648539066314697, + "rewards/margins": -0.0009576906450092793, + "rewards/rejected": -0.0011071630287915468, + "step": 19 + }, + { + "epoch": 0.043907793633369926, + "grad_norm": 1.3268272876739502, + "learning_rate": 1.0000000000000002e-06, + "logits/chosen": -2.305075168609619, + "logits/rejected": -2.3672685623168945, + "logps/chosen": -11.384998321533203, + "logps/rejected": -9.560636520385742, + "loss": 0.9753, + "rewards/accuracies": 0.46875, + "rewards/chosen": 0.0010676621459424496, + "rewards/margins": -0.003087827702984214, + "rewards/rejected": 0.004155490081757307, + "step": 20 + }, + { + "epoch": 0.04610318331503842, + "grad_norm": 1.0563515424728394, + "learning_rate": 1.0500000000000001e-06, + "logits/chosen": -2.2942605018615723, + "logits/rejected": -2.363292694091797, + "logps/chosen": -15.190650939941406, + "logps/rejected": -16.022212982177734, + "loss": 1.1478, + "rewards/accuracies": 0.5, + "rewards/chosen": 3.312993794679642e-05, + "rewards/margins": 0.004593318793922663, + "rewards/rejected": -0.004560189321637154, + "step": 21 + }, + { + "epoch": 0.048298572996706916, + "grad_norm": 1.1235427856445312, + "learning_rate": 1.1e-06, + "logits/chosen": -2.3363523483276367, + "logits/rejected": -2.35919189453125, + "logps/chosen": -13.10991096496582, + "logps/rejected": -12.458131790161133, + "loss": 1.045, + "rewards/accuracies": 0.59375, + "rewards/chosen": 9.548966772854328e-05, + "rewards/margins": -2.5561952497810125e-05, + "rewards/rejected": 0.00012105179484933615, + "step": 22 + }, + { + "epoch": 0.050493962678375415, + "grad_norm": 1.0452659130096436, + "learning_rate": 1.1500000000000002e-06, + "logits/chosen": -2.2859370708465576, + "logits/rejected": -2.472400665283203, + "logps/chosen": -13.347756385803223, + "logps/rejected": -11.171856880187988, + "loss": 1.0528, + "rewards/accuracies": 0.46875, + "rewards/chosen": 0.0030131523963063955, + "rewards/margins": 0.0002964767627418041, + "rewards/rejected": 0.0027166756335645914, + "step": 23 + }, + { + "epoch": 0.052689352360043906, + "grad_norm": 1.1159127950668335, + "learning_rate": 1.2000000000000002e-06, + "logits/chosen": -2.381169319152832, + "logits/rejected": -2.315974712371826, + "logps/chosen": -16.52190399169922, + "logps/rejected": -13.986876487731934, + "loss": 1.1217, + "rewards/accuracies": 0.46875, + "rewards/chosen": -0.002126524457708001, + "rewards/margins": -0.0030577415600419044, + "rewards/rejected": 0.0009312168695032597, + "step": 24 + }, + { + "epoch": 0.054884742041712405, + "grad_norm": 1.2672061920166016, + "learning_rate": 1.25e-06, + "logits/chosen": -2.32135272026062, + "logits/rejected": -2.2802534103393555, + "logps/chosen": -14.381074905395508, + "logps/rejected": -12.72801399230957, + "loss": 1.0435, + "rewards/accuracies": 0.46875, + "rewards/chosen": 0.0030421048868447542, + "rewards/margins": -0.00030010956106707454, + "rewards/rejected": 0.003342214971780777, + "step": 25 + }, + { + "epoch": 0.0570801317233809, + "grad_norm": 1.1465411186218262, + "learning_rate": 1.3e-06, + "logits/chosen": -2.2586517333984375, + "logits/rejected": -2.377082586288452, + "logps/chosen": -11.207740783691406, + "logps/rejected": -13.343524932861328, + "loss": 1.0674, + "rewards/accuracies": 0.53125, + "rewards/chosen": 0.0017482821131125093, + "rewards/margins": 0.002273987280204892, + "rewards/rejected": -0.0005257053999230266, + "step": 26 + }, + { + "epoch": 0.059275521405049394, + "grad_norm": 1.0358684062957764, + "learning_rate": 1.3500000000000002e-06, + "logits/chosen": -2.31711745262146, + "logits/rejected": -2.3493282794952393, + "logps/chosen": -14.618535995483398, + "logps/rejected": -12.928030967712402, + "loss": 1.0852, + "rewards/accuracies": 0.46875, + "rewards/chosen": 0.004235025029629469, + "rewards/margins": -0.0006879566935822368, + "rewards/rejected": 0.00492298137396574, + "step": 27 + }, + { + "epoch": 0.06147091108671789, + "grad_norm": 1.1591858863830566, + "learning_rate": 1.4000000000000001e-06, + "logits/chosen": -2.310506582260132, + "logits/rejected": -2.2369637489318848, + "logps/chosen": -11.844167709350586, + "logps/rejected": -13.115419387817383, + "loss": 1.0665, + "rewards/accuracies": 0.59375, + "rewards/chosen": 0.003957665525376797, + "rewards/margins": 0.007251821458339691, + "rewards/rejected": -0.0032941552344709635, + "step": 28 + }, + { + "epoch": 0.06366630076838639, + "grad_norm": 1.2100194692611694, + "learning_rate": 1.45e-06, + "logits/chosen": -2.384784698486328, + "logits/rejected": -2.348393201828003, + "logps/chosen": -12.626264572143555, + "logps/rejected": -11.345318794250488, + "loss": 0.9866, + "rewards/accuracies": 0.5, + "rewards/chosen": 1.857941970229149e-05, + "rewards/margins": -8.993363007903099e-05, + "rewards/rejected": 0.00010851328261196613, + "step": 29 + }, + { + "epoch": 0.06586169045005488, + "grad_norm": 1.105622410774231, + "learning_rate": 1.5e-06, + "logits/chosen": -2.317814826965332, + "logits/rejected": -2.313504695892334, + "logps/chosen": -14.957558631896973, + "logps/rejected": -13.618270874023438, + "loss": 1.0818, + "rewards/accuracies": 0.5, + "rewards/chosen": 0.0022040181793272495, + "rewards/margins": 0.0015496157575398684, + "rewards/rejected": 0.0006544026546180248, + "step": 30 + }, + { + "epoch": 0.06805708013172337, + "grad_norm": 1.2689664363861084, + "learning_rate": 1.5500000000000002e-06, + "logits/chosen": -2.418194532394409, + "logits/rejected": -2.4411373138427734, + "logps/chosen": -11.708237648010254, + "logps/rejected": -13.883282661437988, + "loss": 1.058, + "rewards/accuracies": 0.46875, + "rewards/chosen": 0.004741518292576075, + "rewards/margins": -0.0009018751443363726, + "rewards/rejected": 0.005643392913043499, + "step": 31 + }, + { + "epoch": 0.07025246981339188, + "grad_norm": 1.3097610473632812, + "learning_rate": 1.6000000000000001e-06, + "logits/chosen": -2.3038244247436523, + "logits/rejected": -2.458163261413574, + "logps/chosen": -12.498970031738281, + "logps/rejected": -10.449270248413086, + "loss": 1.0015, + "rewards/accuracies": 0.53125, + "rewards/chosen": 0.0064272889867424965, + "rewards/margins": 0.009535685181617737, + "rewards/rejected": -0.0031083961948752403, + "step": 32 + }, + { + "epoch": 0.07244785949506037, + "grad_norm": 1.3711282014846802, + "learning_rate": 1.6500000000000003e-06, + "logits/chosen": -2.2958526611328125, + "logits/rejected": -2.357262372970581, + "logps/chosen": -9.87129020690918, + "logps/rejected": -9.644401550292969, + "loss": 0.9782, + "rewards/accuracies": 0.625, + "rewards/chosen": 0.006264886818826199, + "rewards/margins": 0.007479904685169458, + "rewards/rejected": -0.001215018448419869, + "step": 33 + }, + { + "epoch": 0.07464324917672886, + "grad_norm": 1.0014218091964722, + "learning_rate": 1.7000000000000002e-06, + "logits/chosen": -2.3077902793884277, + "logits/rejected": -2.32836651802063, + "logps/chosen": -13.210323333740234, + "logps/rejected": -10.678458213806152, + "loss": 1.0215, + "rewards/accuracies": 0.5625, + "rewards/chosen": 0.003474751953035593, + "rewards/margins": 0.005452706944197416, + "rewards/rejected": -0.0019779547583311796, + "step": 34 + }, + { + "epoch": 0.07683863885839737, + "grad_norm": 1.5878692865371704, + "learning_rate": 1.75e-06, + "logits/chosen": -2.3073747158050537, + "logits/rejected": -2.2542941570281982, + "logps/chosen": -11.676526069641113, + "logps/rejected": -12.513848304748535, + "loss": 1.0121, + "rewards/accuracies": 0.46875, + "rewards/chosen": 0.0010510723805055022, + "rewards/margins": -0.0036773495376110077, + "rewards/rejected": 0.004728421568870544, + "step": 35 + }, + { + "epoch": 0.07903402854006586, + "grad_norm": 1.313846468925476, + "learning_rate": 1.8000000000000001e-06, + "logits/chosen": -2.402480125427246, + "logits/rejected": -2.351649284362793, + "logps/chosen": -12.576855659484863, + "logps/rejected": -11.105303764343262, + "loss": 0.9857, + "rewards/accuracies": 0.5, + "rewards/chosen": -0.0032139169052243233, + "rewards/margins": -0.005750270560383797, + "rewards/rejected": 0.0025363534223288298, + "step": 36 + }, + { + "epoch": 0.08122941822173436, + "grad_norm": 1.3616929054260254, + "learning_rate": 1.85e-06, + "logits/chosen": -2.2891414165496826, + "logits/rejected": -2.3991661071777344, + "logps/chosen": -14.187095642089844, + "logps/rejected": -11.036577224731445, + "loss": 0.9868, + "rewards/accuracies": 0.625, + "rewards/chosen": 0.006025272887200117, + "rewards/margins": 0.008252888917922974, + "rewards/rejected": -0.0022276148665696383, + "step": 37 + }, + { + "epoch": 0.08342480790340286, + "grad_norm": 1.3414860963821411, + "learning_rate": 1.9000000000000002e-06, + "logits/chosen": -2.377753973007202, + "logits/rejected": -2.4487216472625732, + "logps/chosen": -18.255245208740234, + "logps/rejected": -12.83712387084961, + "loss": 1.0333, + "rewards/accuracies": 0.59375, + "rewards/chosen": 0.0022521866485476494, + "rewards/margins": 0.004233865533024073, + "rewards/rejected": -0.001981678418815136, + "step": 38 + }, + { + "epoch": 0.08562019758507135, + "grad_norm": 1.0859737396240234, + "learning_rate": 1.9500000000000004e-06, + "logits/chosen": -2.312309980392456, + "logits/rejected": -2.2821502685546875, + "logps/chosen": -14.578563690185547, + "logps/rejected": -15.732474327087402, + "loss": 1.0939, + "rewards/accuracies": 0.46875, + "rewards/chosen": 0.0022968475241214037, + "rewards/margins": 0.0003764451830647886, + "rewards/rejected": 0.0019204019336029887, + "step": 39 + }, + { + "epoch": 0.08781558726673985, + "grad_norm": 1.2043111324310303, + "learning_rate": 2.0000000000000003e-06, + "logits/chosen": -2.3180580139160156, + "logits/rejected": -2.3157169818878174, + "logps/chosen": -12.22863483428955, + "logps/rejected": -11.449076652526855, + "loss": 1.0204, + "rewards/accuracies": 0.40625, + "rewards/chosen": 0.00017814873717725277, + "rewards/margins": -0.004647126421332359, + "rewards/rejected": 0.004825274925678968, + "step": 40 + }, + { + "epoch": 0.09001097694840834, + "grad_norm": 1.2361137866973877, + "learning_rate": 2.05e-06, + "logits/chosen": -2.343757152557373, + "logits/rejected": -2.3358938694000244, + "logps/chosen": -12.45596694946289, + "logps/rejected": -12.058631896972656, + "loss": 1.0172, + "rewards/accuracies": 0.40625, + "rewards/chosen": 0.0018289226572960615, + "rewards/margins": 0.000773195642977953, + "rewards/rejected": 0.0010557263158261776, + "step": 41 + }, + { + "epoch": 0.09220636663007684, + "grad_norm": 1.1400678157806396, + "learning_rate": 2.1000000000000002e-06, + "logits/chosen": -2.3396897315979004, + "logits/rejected": -2.262922763824463, + "logps/chosen": -11.612154006958008, + "logps/rejected": -10.86433219909668, + "loss": 0.9884, + "rewards/accuracies": 0.5, + "rewards/chosen": 0.003359719179570675, + "rewards/margins": 0.0006888797506690025, + "rewards/rejected": 0.0026708384975790977, + "step": 42 + }, + { + "epoch": 0.09440175631174534, + "grad_norm": 1.0593880414962769, + "learning_rate": 2.15e-06, + "logits/chosen": -2.27569580078125, + "logits/rejected": -2.345578670501709, + "logps/chosen": -16.460983276367188, + "logps/rejected": -10.75358772277832, + "loss": 1.006, + "rewards/accuracies": 0.75, + "rewards/chosen": 0.008105424232780933, + "rewards/margins": 0.011831814423203468, + "rewards/rejected": -0.003726390190422535, + "step": 43 + }, + { + "epoch": 0.09659714599341383, + "grad_norm": 1.3210548162460327, + "learning_rate": 2.2e-06, + "logits/chosen": -2.3760862350463867, + "logits/rejected": -2.2515220642089844, + "logps/chosen": -14.993949890136719, + "logps/rejected": -13.207966804504395, + "loss": 1.0259, + "rewards/accuracies": 0.46875, + "rewards/chosen": 0.00593589898198843, + "rewards/margins": 0.00230758311226964, + "rewards/rejected": 0.0036283163353800774, + "step": 44 + }, + { + "epoch": 0.09879253567508232, + "grad_norm": 0.9838950634002686, + "learning_rate": 2.25e-06, + "logits/chosen": -2.3689630031585693, + "logits/rejected": -2.3999180793762207, + "logps/chosen": -13.113698959350586, + "logps/rejected": -11.06283950805664, + "loss": 0.9872, + "rewards/accuracies": 0.75, + "rewards/chosen": 0.008952252566814423, + "rewards/margins": 0.010060377418994904, + "rewards/rejected": -0.0011081250850111246, + "step": 45 + }, + { + "epoch": 0.10098792535675083, + "grad_norm": 1.2060089111328125, + "learning_rate": 2.3000000000000004e-06, + "logits/chosen": -2.3371407985687256, + "logits/rejected": -2.3095602989196777, + "logps/chosen": -10.346152305603027, + "logps/rejected": -10.529175758361816, + "loss": 0.9568, + "rewards/accuracies": 0.5625, + "rewards/chosen": 0.005232983268797398, + "rewards/margins": 0.006484426558017731, + "rewards/rejected": -0.0012514438712969422, + "step": 46 + }, + { + "epoch": 0.10318331503841932, + "grad_norm": 1.1937215328216553, + "learning_rate": 2.35e-06, + "logits/chosen": -2.383635997772217, + "logits/rejected": -2.251980781555176, + "logps/chosen": -14.678833961486816, + "logps/rejected": -14.274091720581055, + "loss": 1.1045, + "rewards/accuracies": 0.59375, + "rewards/chosen": -0.0001039394992403686, + "rewards/margins": 0.0014265230856835842, + "rewards/rejected": -0.00153046241030097, + "step": 47 + }, + { + "epoch": 0.10537870472008781, + "grad_norm": 1.2106831073760986, + "learning_rate": 2.4000000000000003e-06, + "logits/chosen": -2.307473659515381, + "logits/rejected": -2.270702362060547, + "logps/chosen": -14.072576522827148, + "logps/rejected": -10.658452987670898, + "loss": 1.0071, + "rewards/accuracies": 0.4375, + "rewards/chosen": 0.0023170302156358957, + "rewards/margins": 0.0011877454817295074, + "rewards/rejected": 0.0011292845010757446, + "step": 48 + }, + { + "epoch": 0.10757409440175632, + "grad_norm": 1.1354262828826904, + "learning_rate": 2.4500000000000003e-06, + "logits/chosen": -2.248020887374878, + "logits/rejected": -2.2451305389404297, + "logps/chosen": -15.853403091430664, + "logps/rejected": -15.368898391723633, + "loss": 1.1177, + "rewards/accuracies": 0.53125, + "rewards/chosen": 0.0036325519904494286, + "rewards/margins": 0.00015486334450542927, + "rewards/rejected": 0.0034776891116052866, + "step": 49 + }, + { + "epoch": 0.10976948408342481, + "grad_norm": 1.0705238580703735, + "learning_rate": 2.5e-06, + "logits/chosen": -2.4041616916656494, + "logits/rejected": -2.383108139038086, + "logps/chosen": -18.875839233398438, + "logps/rejected": -12.37950325012207, + "loss": 1.0141, + "rewards/accuracies": 0.46875, + "rewards/chosen": 0.003123737871646881, + "rewards/margins": -0.0004079572972841561, + "rewards/rejected": 0.00353169534355402, + "step": 50 + }, + { + "epoch": 0.1119648737650933, + "grad_norm": 1.26189386844635, + "learning_rate": 2.55e-06, + "logits/chosen": -2.3076915740966797, + "logits/rejected": -2.283825397491455, + "logps/chosen": -13.641839027404785, + "logps/rejected": -11.214380264282227, + "loss": 0.9837, + "rewards/accuracies": 0.46875, + "rewards/chosen": 0.002529280027374625, + "rewards/margins": 8.014822378754616e-05, + "rewards/rejected": 0.002449131803587079, + "step": 51 + }, + { + "epoch": 0.1141602634467618, + "grad_norm": 1.302383542060852, + "learning_rate": 2.6e-06, + "logits/chosen": -2.2420313358306885, + "logits/rejected": -2.3660824298858643, + "logps/chosen": -11.503904342651367, + "logps/rejected": -11.648876190185547, + "loss": 1.0438, + "rewards/accuracies": 0.53125, + "rewards/chosen": -0.000531287572812289, + "rewards/margins": 0.0005554634844884276, + "rewards/rejected": -0.001086751464754343, + "step": 52 + }, + { + "epoch": 0.1163556531284303, + "grad_norm": 1.2518279552459717, + "learning_rate": 2.6500000000000005e-06, + "logits/chosen": -2.3373780250549316, + "logits/rejected": -2.279087543487549, + "logps/chosen": -16.882110595703125, + "logps/rejected": -15.901355743408203, + "loss": 1.1307, + "rewards/accuracies": 0.59375, + "rewards/chosen": 0.0010501868091523647, + "rewards/margins": 0.00016966351540759206, + "rewards/rejected": 0.0008805230027064681, + "step": 53 + }, + { + "epoch": 0.11855104281009879, + "grad_norm": 1.1796338558197021, + "learning_rate": 2.7000000000000004e-06, + "logits/chosen": -2.3147006034851074, + "logits/rejected": -2.245007276535034, + "logps/chosen": -13.884672164916992, + "logps/rejected": -12.487886428833008, + "loss": 1.0253, + "rewards/accuracies": 0.5625, + "rewards/chosen": 0.007091285195201635, + "rewards/margins": 0.004812297876924276, + "rewards/rejected": 0.002278987318277359, + "step": 54 + }, + { + "epoch": 0.1207464324917673, + "grad_norm": 1.2227309942245483, + "learning_rate": 2.7500000000000004e-06, + "logits/chosen": -2.2176713943481445, + "logits/rejected": -2.2463364601135254, + "logps/chosen": -13.619998931884766, + "logps/rejected": -12.540321350097656, + "loss": 1.0239, + "rewards/accuracies": 0.375, + "rewards/chosen": -0.0013588806614279747, + "rewards/margins": -0.004556029103696346, + "rewards/rejected": 0.0031971491407603025, + "step": 55 + }, + { + "epoch": 0.12294182217343579, + "grad_norm": 0.9918203949928284, + "learning_rate": 2.8000000000000003e-06, + "logits/chosen": -2.3580827713012695, + "logits/rejected": -2.3567821979522705, + "logps/chosen": -12.459843635559082, + "logps/rejected": -9.410993576049805, + "loss": 0.9476, + "rewards/accuracies": 0.5625, + "rewards/chosen": 0.005544860381633043, + "rewards/margins": 0.0024246471002697945, + "rewards/rejected": 0.003120213747024536, + "step": 56 + }, + { + "epoch": 0.1251372118551043, + "grad_norm": 1.6529310941696167, + "learning_rate": 2.85e-06, + "logits/chosen": -2.3447060585021973, + "logits/rejected": -2.249061107635498, + "logps/chosen": -11.913742065429688, + "logps/rejected": -11.23048210144043, + "loss": 0.9893, + "rewards/accuracies": 0.5625, + "rewards/chosen": 0.004630698822438717, + "rewards/margins": 0.005256508477032185, + "rewards/rejected": -0.0006258084904402494, + "step": 57 + }, + { + "epoch": 0.12733260153677278, + "grad_norm": 1.0494027137756348, + "learning_rate": 2.9e-06, + "logits/chosen": -2.2313132286071777, + "logits/rejected": -2.401317596435547, + "logps/chosen": -18.439748764038086, + "logps/rejected": -11.610260009765625, + "loss": 1.0282, + "rewards/accuracies": 0.625, + "rewards/chosen": 0.0071330079808831215, + "rewards/margins": 0.006482796743512154, + "rewards/rejected": 0.0006502112373709679, + "step": 58 + }, + { + "epoch": 0.12952799121844127, + "grad_norm": 1.4393824338912964, + "learning_rate": 2.95e-06, + "logits/chosen": -2.3876466751098633, + "logits/rejected": -2.399052143096924, + "logps/chosen": -14.200356483459473, + "logps/rejected": -13.737493515014648, + "loss": 1.0595, + "rewards/accuracies": 0.5, + "rewards/chosen": 0.001972971251234412, + "rewards/margins": 0.002361137419939041, + "rewards/rejected": -0.0003881664015352726, + "step": 59 + }, + { + "epoch": 0.13172338090010977, + "grad_norm": 1.3774553537368774, + "learning_rate": 3e-06, + "logits/chosen": -2.210278034210205, + "logits/rejected": -2.2798633575439453, + "logps/chosen": -13.26652717590332, + "logps/rejected": -11.69228458404541, + "loss": 0.9983, + "rewards/accuracies": 0.5, + "rewards/chosen": 0.0036805993877351284, + "rewards/margins": 0.0023272125981748104, + "rewards/rejected": 0.001353386789560318, + "step": 60 + }, + { + "epoch": 0.13391877058177826, + "grad_norm": 1.3280311822891235, + "learning_rate": 3.05e-06, + "logits/chosen": -2.2844367027282715, + "logits/rejected": -2.416640281677246, + "logps/chosen": -13.028985023498535, + "logps/rejected": -10.592105865478516, + "loss": 0.984, + "rewards/accuracies": 0.65625, + "rewards/chosen": 0.00871636625379324, + "rewards/margins": 0.01077802013605833, + "rewards/rejected": -0.0020616534166038036, + "step": 61 + }, + { + "epoch": 0.13611416026344675, + "grad_norm": 1.5957064628601074, + "learning_rate": 3.1000000000000004e-06, + "logits/chosen": -2.2388463020324707, + "logits/rejected": -2.429072856903076, + "logps/chosen": -16.77328109741211, + "logps/rejected": -11.056230545043945, + "loss": 0.9916, + "rewards/accuracies": 0.75, + "rewards/chosen": 0.011861161328852177, + "rewards/margins": 0.015179607085883617, + "rewards/rejected": -0.003318445524200797, + "step": 62 + }, + { + "epoch": 0.13830954994511527, + "grad_norm": 1.3382227420806885, + "learning_rate": 3.1500000000000003e-06, + "logits/chosen": -2.3019542694091797, + "logits/rejected": -2.380207061767578, + "logps/chosen": -15.527379035949707, + "logps/rejected": -13.00551986694336, + "loss": 1.0461, + "rewards/accuracies": 0.625, + "rewards/chosen": 0.014199727214872837, + "rewards/margins": 0.004217819310724735, + "rewards/rejected": 0.009981908835470676, + "step": 63 + }, + { + "epoch": 0.14050493962678376, + "grad_norm": 1.4465454816818237, + "learning_rate": 3.2000000000000003e-06, + "logits/chosen": -2.293320417404175, + "logits/rejected": -2.3460850715637207, + "logps/chosen": -12.336854934692383, + "logps/rejected": -12.035362243652344, + "loss": 1.0247, + "rewards/accuracies": 0.40625, + "rewards/chosen": 0.0019360918086022139, + "rewards/margins": -0.004731356166303158, + "rewards/rejected": 0.006667448207736015, + "step": 64 + }, + { + "epoch": 0.14270032930845225, + "grad_norm": 1.438913106918335, + "learning_rate": 3.2500000000000002e-06, + "logits/chosen": -2.356405258178711, + "logits/rejected": -2.381514072418213, + "logps/chosen": -16.52509117126465, + "logps/rejected": -11.303572654724121, + "loss": 1.0138, + "rewards/accuracies": 0.6875, + "rewards/chosen": 0.012714626267552376, + "rewards/margins": 0.010571034625172615, + "rewards/rejected": 0.0021435916423797607, + "step": 65 + }, + { + "epoch": 0.14489571899012074, + "grad_norm": 1.352678656578064, + "learning_rate": 3.3000000000000006e-06, + "logits/chosen": -2.3136496543884277, + "logits/rejected": -2.3507792949676514, + "logps/chosen": -13.596353530883789, + "logps/rejected": -9.79554557800293, + "loss": 0.9822, + "rewards/accuracies": 0.625, + "rewards/chosen": 0.012121602892875671, + "rewards/margins": 0.015090488828718662, + "rewards/rejected": -0.0029688861686736345, + "step": 66 + }, + { + "epoch": 0.14709110867178923, + "grad_norm": 1.079323410987854, + "learning_rate": 3.3500000000000005e-06, + "logits/chosen": -2.3352060317993164, + "logits/rejected": -2.3463046550750732, + "logps/chosen": -12.48592758178711, + "logps/rejected": -11.822932243347168, + "loss": 1.0085, + "rewards/accuracies": 0.5625, + "rewards/chosen": 0.008349809795618057, + "rewards/margins": -0.00189165398478508, + "rewards/rejected": 0.010241463780403137, + "step": 67 + }, + { + "epoch": 0.14928649835345773, + "grad_norm": 1.618390440940857, + "learning_rate": 3.4000000000000005e-06, + "logits/chosen": -2.333874225616455, + "logits/rejected": -2.4121692180633545, + "logps/chosen": -14.930086135864258, + "logps/rejected": -11.14267349243164, + "loss": 0.9708, + "rewards/accuracies": 0.625, + "rewards/chosen": 0.015850193798542023, + "rewards/margins": 0.010664023458957672, + "rewards/rejected": 0.005186168942600489, + "step": 68 + }, + { + "epoch": 0.15148188803512624, + "grad_norm": 1.1727185249328613, + "learning_rate": 3.45e-06, + "logits/chosen": -2.327249765396118, + "logits/rejected": -2.3228392601013184, + "logps/chosen": -17.513296127319336, + "logps/rejected": -11.67719841003418, + "loss": 1.0356, + "rewards/accuracies": 0.4375, + "rewards/chosen": 0.00726369908079505, + "rewards/margins": -0.003262670710682869, + "rewards/rejected": 0.010526370257139206, + "step": 69 + }, + { + "epoch": 0.15367727771679474, + "grad_norm": 1.5009926557540894, + "learning_rate": 3.5e-06, + "logits/chosen": -2.381566047668457, + "logits/rejected": -2.2929749488830566, + "logps/chosen": -13.011417388916016, + "logps/rejected": -13.374524116516113, + "loss": 1.0615, + "rewards/accuracies": 0.46875, + "rewards/chosen": 0.00458131218329072, + "rewards/margins": 0.0018186615779995918, + "rewards/rejected": 0.002762650139629841, + "step": 70 + }, + { + "epoch": 0.15587266739846323, + "grad_norm": 1.0766772031784058, + "learning_rate": 3.5500000000000003e-06, + "logits/chosen": -2.3763010501861572, + "logits/rejected": -2.3684229850769043, + "logps/chosen": -12.743946075439453, + "logps/rejected": -9.790771484375, + "loss": 0.985, + "rewards/accuracies": 0.53125, + "rewards/chosen": 0.00604588445276022, + "rewards/margins": -0.0017442656680941582, + "rewards/rejected": 0.007790150120854378, + "step": 71 + }, + { + "epoch": 0.15806805708013172, + "grad_norm": 1.3915514945983887, + "learning_rate": 3.6000000000000003e-06, + "logits/chosen": -2.255495071411133, + "logits/rejected": -2.207207679748535, + "logps/chosen": -13.701947212219238, + "logps/rejected": -15.165384292602539, + "loss": 1.0743, + "rewards/accuracies": 0.6875, + "rewards/chosen": 0.011631378903985023, + "rewards/margins": 0.005737782455980778, + "rewards/rejected": 0.00589359737932682, + "step": 72 + }, + { + "epoch": 0.1602634467618002, + "grad_norm": 1.7364999055862427, + "learning_rate": 3.65e-06, + "logits/chosen": -2.413259506225586, + "logits/rejected": -2.469926357269287, + "logps/chosen": -17.120332717895508, + "logps/rejected": -10.822717666625977, + "loss": 0.9883, + "rewards/accuracies": 0.59375, + "rewards/chosen": 0.017608050256967545, + "rewards/margins": 0.004501630552113056, + "rewards/rejected": 0.013106418773531914, + "step": 73 + }, + { + "epoch": 0.16245883644346873, + "grad_norm": 1.3227636814117432, + "learning_rate": 3.7e-06, + "logits/chosen": -2.308757781982422, + "logits/rejected": -2.3778414726257324, + "logps/chosen": -15.324178695678711, + "logps/rejected": -9.740997314453125, + "loss": 0.9552, + "rewards/accuracies": 0.53125, + "rewards/chosen": 0.008803250268101692, + "rewards/margins": -0.003404664108529687, + "rewards/rejected": 0.012207916006445885, + "step": 74 + }, + { + "epoch": 0.16465422612513722, + "grad_norm": 1.4592341184616089, + "learning_rate": 3.7500000000000005e-06, + "logits/chosen": -2.389984130859375, + "logits/rejected": -2.463046073913574, + "logps/chosen": -14.928573608398438, + "logps/rejected": -10.570882797241211, + "loss": 0.9779, + "rewards/accuracies": 0.625, + "rewards/chosen": 0.01354161836206913, + "rewards/margins": 0.005444116424769163, + "rewards/rejected": 0.00809750147163868, + "step": 75 + }, + { + "epoch": 0.1668496158068057, + "grad_norm": 1.2499895095825195, + "learning_rate": 3.8000000000000005e-06, + "logits/chosen": -2.319549322128296, + "logits/rejected": -2.4380674362182617, + "logps/chosen": -12.940937995910645, + "logps/rejected": -10.083399772644043, + "loss": 0.9944, + "rewards/accuracies": 0.46875, + "rewards/chosen": 0.009357284754514694, + "rewards/margins": -0.006600930355489254, + "rewards/rejected": 0.015958216041326523, + "step": 76 + }, + { + "epoch": 0.1690450054884742, + "grad_norm": 1.7567248344421387, + "learning_rate": 3.85e-06, + "logits/chosen": -2.393270969390869, + "logits/rejected": -2.2500927448272705, + "logps/chosen": -11.881815910339355, + "logps/rejected": -12.437444686889648, + "loss": 1.0173, + "rewards/accuracies": 0.5625, + "rewards/chosen": 0.02429443597793579, + "rewards/margins": 0.014743359759449959, + "rewards/rejected": 0.009551076218485832, + "step": 77 + }, + { + "epoch": 0.1712403951701427, + "grad_norm": 1.4113529920578003, + "learning_rate": 3.900000000000001e-06, + "logits/chosen": -2.3604631423950195, + "logits/rejected": -2.360464334487915, + "logps/chosen": -13.314837455749512, + "logps/rejected": -10.539698600769043, + "loss": 0.9734, + "rewards/accuracies": 0.5, + "rewards/chosen": 0.012358862906694412, + "rewards/margins": -0.004256892018020153, + "rewards/rejected": 0.01661575585603714, + "step": 78 + }, + { + "epoch": 0.1734357848518112, + "grad_norm": 1.5121917724609375, + "learning_rate": 3.95e-06, + "logits/chosen": -2.3660640716552734, + "logits/rejected": -2.3211538791656494, + "logps/chosen": -13.57307243347168, + "logps/rejected": -11.361422538757324, + "loss": 1.0219, + "rewards/accuracies": 0.375, + "rewards/chosen": 0.0046729762107133865, + "rewards/margins": -0.007223246619105339, + "rewards/rejected": 0.011896222829818726, + "step": 79 + }, + { + "epoch": 0.1756311745334797, + "grad_norm": 1.6794378757476807, + "learning_rate": 4.000000000000001e-06, + "logits/chosen": -2.333510398864746, + "logits/rejected": -2.325746774673462, + "logps/chosen": -12.330310821533203, + "logps/rejected": -13.084382057189941, + "loss": 1.0129, + "rewards/accuracies": 0.4375, + "rewards/chosen": 0.015081634745001793, + "rewards/margins": -0.00490574212744832, + "rewards/rejected": 0.019987378269433975, + "step": 80 + }, + { + "epoch": 0.1778265642151482, + "grad_norm": 1.487237572669983, + "learning_rate": 4.05e-06, + "logits/chosen": -2.41157865524292, + "logits/rejected": -2.300631523132324, + "logps/chosen": -13.241340637207031, + "logps/rejected": -13.954183578491211, + "loss": 1.0723, + "rewards/accuracies": 0.59375, + "rewards/chosen": 0.022465569898486137, + "rewards/margins": 0.0036638593301177025, + "rewards/rejected": 0.01880171149969101, + "step": 81 + }, + { + "epoch": 0.1800219538968167, + "grad_norm": 1.2064540386199951, + "learning_rate": 4.1e-06, + "logits/chosen": -2.2773404121398926, + "logits/rejected": -2.4026074409484863, + "logps/chosen": -16.395071029663086, + "logps/rejected": -11.761744499206543, + "loss": 1.049, + "rewards/accuracies": 0.6875, + "rewards/chosen": 0.010554658249020576, + "rewards/margins": 0.014174232259392738, + "rewards/rejected": -0.003619574476033449, + "step": 82 + }, + { + "epoch": 0.18221734357848518, + "grad_norm": 1.4717923402786255, + "learning_rate": 4.15e-06, + "logits/chosen": -2.3314809799194336, + "logits/rejected": -2.4027044773101807, + "logps/chosen": -13.08586311340332, + "logps/rejected": -12.964557647705078, + "loss": 1.0318, + "rewards/accuracies": 0.46875, + "rewards/chosen": 0.022713469341397285, + "rewards/margins": 0.01534382812678814, + "rewards/rejected": 0.007369642611593008, + "step": 83 + }, + { + "epoch": 0.18441273326015367, + "grad_norm": 1.2905617952346802, + "learning_rate": 4.2000000000000004e-06, + "logits/chosen": -2.3403005599975586, + "logits/rejected": -2.3712852001190186, + "logps/chosen": -13.521167755126953, + "logps/rejected": -9.760311126708984, + "loss": 0.963, + "rewards/accuracies": 0.59375, + "rewards/chosen": 0.026714814826846123, + "rewards/margins": 0.017565395683050156, + "rewards/rejected": 0.009149417281150818, + "step": 84 + }, + { + "epoch": 0.18660812294182216, + "grad_norm": 1.4235241413116455, + "learning_rate": 4.25e-06, + "logits/chosen": -2.3429999351501465, + "logits/rejected": -2.3919286727905273, + "logps/chosen": -12.980363845825195, + "logps/rejected": -10.314764976501465, + "loss": 0.9704, + "rewards/accuracies": 0.6875, + "rewards/chosen": 0.04204602539539337, + "rewards/margins": 0.01917717233300209, + "rewards/rejected": 0.022868849337100983, + "step": 85 + }, + { + "epoch": 0.18880351262349068, + "grad_norm": 1.1854406595230103, + "learning_rate": 4.3e-06, + "logits/chosen": -2.269547939300537, + "logits/rejected": -2.2928738594055176, + "logps/chosen": -13.91822624206543, + "logps/rejected": -11.471654891967773, + "loss": 0.9748, + "rewards/accuracies": 0.625, + "rewards/chosen": 0.03401780128479004, + "rewards/margins": 0.021458495408296585, + "rewards/rejected": 0.012559305876493454, + "step": 86 + }, + { + "epoch": 0.19099890230515917, + "grad_norm": 1.9034478664398193, + "learning_rate": 4.350000000000001e-06, + "logits/chosen": -2.300184488296509, + "logits/rejected": -2.249917984008789, + "logps/chosen": -11.432275772094727, + "logps/rejected": -12.034334182739258, + "loss": 0.996, + "rewards/accuracies": 0.625, + "rewards/chosen": 0.03009907528758049, + "rewards/margins": 0.015358498319983482, + "rewards/rejected": 0.014740575104951859, + "step": 87 + }, + { + "epoch": 0.19319429198682767, + "grad_norm": 2.1222307682037354, + "learning_rate": 4.4e-06, + "logits/chosen": -2.331315517425537, + "logits/rejected": -2.329355478286743, + "logps/chosen": -15.621415138244629, + "logps/rejected": -11.350690841674805, + "loss": 0.9767, + "rewards/accuracies": 0.5625, + "rewards/chosen": 0.044887036085128784, + "rewards/margins": 0.01875218376517296, + "rewards/rejected": 0.026134852319955826, + "step": 88 + }, + { + "epoch": 0.19538968166849616, + "grad_norm": 1.4220390319824219, + "learning_rate": 4.450000000000001e-06, + "logits/chosen": -2.336406707763672, + "logits/rejected": -2.3903348445892334, + "logps/chosen": -15.872965812683105, + "logps/rejected": -12.38022232055664, + "loss": 1.0457, + "rewards/accuracies": 0.5625, + "rewards/chosen": 0.01367251481860876, + "rewards/margins": -0.001530464505776763, + "rewards/rejected": 0.015202978625893593, + "step": 89 + }, + { + "epoch": 0.19758507135016465, + "grad_norm": 1.282646656036377, + "learning_rate": 4.5e-06, + "logits/chosen": -2.361445903778076, + "logits/rejected": -2.476417064666748, + "logps/chosen": -14.19363021850586, + "logps/rejected": -15.457003593444824, + "loss": 1.1384, + "rewards/accuracies": 0.5, + "rewards/chosen": 0.02910601533949375, + "rewards/margins": -0.002174281282350421, + "rewards/rejected": 0.03128029778599739, + "step": 90 + }, + { + "epoch": 0.19978046103183314, + "grad_norm": 1.2941416501998901, + "learning_rate": 4.5500000000000005e-06, + "logits/chosen": -2.299647331237793, + "logits/rejected": -2.2841219902038574, + "logps/chosen": -12.112016677856445, + "logps/rejected": -10.684823989868164, + "loss": 0.9786, + "rewards/accuracies": 0.5625, + "rewards/chosen": 0.041696906089782715, + "rewards/margins": 0.029532143846154213, + "rewards/rejected": 0.012164760380983353, + "step": 91 + }, + { + "epoch": 0.20197585071350166, + "grad_norm": 2.0074691772460938, + "learning_rate": 4.600000000000001e-06, + "logits/chosen": -2.437091588973999, + "logits/rejected": -2.2995917797088623, + "logps/chosen": -9.002144813537598, + "logps/rejected": -13.91781234741211, + "loss": 1.0479, + "rewards/accuracies": 0.59375, + "rewards/chosen": 0.03345324099063873, + "rewards/margins": 0.010070934891700745, + "rewards/rejected": 0.023382307961583138, + "step": 92 + }, + { + "epoch": 0.20417124039517015, + "grad_norm": 1.4986943006515503, + "learning_rate": 4.65e-06, + "logits/chosen": -2.3514928817749023, + "logits/rejected": -2.3139781951904297, + "logps/chosen": -13.505949020385742, + "logps/rejected": -12.242940902709961, + "loss": 1.0188, + "rewards/accuracies": 0.5625, + "rewards/chosen": 0.03294369950890541, + "rewards/margins": 0.011969504877924919, + "rewards/rejected": 0.020974192768335342, + "step": 93 + }, + { + "epoch": 0.20636663007683864, + "grad_norm": 1.2458827495574951, + "learning_rate": 4.7e-06, + "logits/chosen": -2.314868927001953, + "logits/rejected": -2.3728139400482178, + "logps/chosen": -11.619760513305664, + "logps/rejected": -11.497888565063477, + "loss": 0.9931, + "rewards/accuracies": 0.59375, + "rewards/chosen": 0.02766992524266243, + "rewards/margins": 0.009395633824169636, + "rewards/rejected": 0.01827428862452507, + "step": 94 + }, + { + "epoch": 0.20856201975850713, + "grad_norm": 1.6697496175765991, + "learning_rate": 4.75e-06, + "logits/chosen": -2.3906478881835938, + "logits/rejected": -2.3166065216064453, + "logps/chosen": -13.515888214111328, + "logps/rejected": -9.598672866821289, + "loss": 0.9525, + "rewards/accuracies": 0.59375, + "rewards/chosen": 0.047668956220149994, + "rewards/margins": 0.002992257010191679, + "rewards/rejected": 0.04467669874429703, + "step": 95 + }, + { + "epoch": 0.21075740944017562, + "grad_norm": 1.3917086124420166, + "learning_rate": 4.800000000000001e-06, + "logits/chosen": -2.2934889793395996, + "logits/rejected": -2.2676901817321777, + "logps/chosen": -13.526961326599121, + "logps/rejected": -11.483183860778809, + "loss": 0.9959, + "rewards/accuracies": 0.59375, + "rewards/chosen": 0.04153824597597122, + "rewards/margins": 0.022040612995624542, + "rewards/rejected": 0.01949763298034668, + "step": 96 + }, + { + "epoch": 0.21295279912184412, + "grad_norm": 1.4579229354858398, + "learning_rate": 4.85e-06, + "logits/chosen": -2.30719256401062, + "logits/rejected": -2.408965826034546, + "logps/chosen": -11.994451522827148, + "logps/rejected": -10.733743667602539, + "loss": 0.9979, + "rewards/accuracies": 0.625, + "rewards/chosen": 0.04512687399983406, + "rewards/margins": -0.0012154264841228724, + "rewards/rejected": 0.04634230211377144, + "step": 97 + }, + { + "epoch": 0.21514818880351264, + "grad_norm": 1.4664539098739624, + "learning_rate": 4.9000000000000005e-06, + "logits/chosen": -2.3413281440734863, + "logits/rejected": -2.3405394554138184, + "logps/chosen": -13.540508270263672, + "logps/rejected": -12.272972106933594, + "loss": 1.0106, + "rewards/accuracies": 0.5, + "rewards/chosen": 0.0718052089214325, + "rewards/margins": 0.014313442632555962, + "rewards/rejected": 0.057491764426231384, + "step": 98 + }, + { + "epoch": 0.21734357848518113, + "grad_norm": 1.4156373739242554, + "learning_rate": 4.95e-06, + "logits/chosen": -2.3612160682678223, + "logits/rejected": -2.297560691833496, + "logps/chosen": -14.15035343170166, + "logps/rejected": -11.349997520446777, + "loss": 1.0359, + "rewards/accuracies": 0.4375, + "rewards/chosen": 0.03496279940009117, + "rewards/margins": -0.026759572327136993, + "rewards/rejected": 0.06172237545251846, + "step": 99 + }, + { + "epoch": 0.21953896816684962, + "grad_norm": 1.8833767175674438, + "learning_rate": 5e-06, + "logits/chosen": -2.3501124382019043, + "logits/rejected": -2.3347649574279785, + "logps/chosen": -9.56277084350586, + "logps/rejected": -8.914642333984375, + "loss": 0.9279, + "rewards/accuracies": 0.6875, + "rewards/chosen": 0.06373147666454315, + "rewards/margins": 0.04788838326931, + "rewards/rejected": 0.015843091532588005, + "step": 100 + }, + { + "epoch": 0.2217343578485181, + "grad_norm": 1.5216327905654907, + "learning_rate": 4.999981196479141e-06, + "logits/chosen": -2.28043794631958, + "logits/rejected": -2.2953336238861084, + "logps/chosen": -16.370956420898438, + "logps/rejected": -14.55150032043457, + "loss": 1.0965, + "rewards/accuracies": 0.53125, + "rewards/chosen": 0.05121631175279617, + "rewards/margins": 0.013376165181398392, + "rewards/rejected": 0.03784014284610748, + "step": 101 + }, + { + "epoch": 0.2239297475301866, + "grad_norm": 1.3872263431549072, + "learning_rate": 4.99992478619942e-06, + "logits/chosen": -2.279304265975952, + "logits/rejected": -2.268181800842285, + "logps/chosen": -15.490918159484863, + "logps/rejected": -13.121916770935059, + "loss": 1.0741, + "rewards/accuracies": 0.625, + "rewards/chosen": 0.05697057023644447, + "rewards/margins": -0.002729692729189992, + "rewards/rejected": 0.059700265526771545, + "step": 102 + }, + { + "epoch": 0.2261251372118551, + "grad_norm": 1.5503108501434326, + "learning_rate": 4.999830770009406e-06, + "logits/chosen": -2.27740478515625, + "logits/rejected": -2.2930843830108643, + "logps/chosen": -11.542291641235352, + "logps/rejected": -10.724907875061035, + "loss": 0.9791, + "rewards/accuracies": 0.625, + "rewards/chosen": 0.03553184121847153, + "rewards/margins": 0.014793504029512405, + "rewards/rejected": 0.020738335326313972, + "step": 103 + }, + { + "epoch": 0.2283205268935236, + "grad_norm": 1.5652024745941162, + "learning_rate": 4.999699149323369e-06, + "logits/chosen": -2.307811737060547, + "logits/rejected": -2.338045597076416, + "logps/chosen": -14.059741973876953, + "logps/rejected": -12.458868026733398, + "loss": 1.0213, + "rewards/accuracies": 0.6875, + "rewards/chosen": 0.08312933892011642, + "rewards/margins": 0.034912753850221634, + "rewards/rejected": 0.04821658134460449, + "step": 104 + }, + { + "epoch": 0.2305159165751921, + "grad_norm": 1.4373472929000854, + "learning_rate": 4.999529926121254e-06, + "logits/chosen": -2.3321356773376465, + "logits/rejected": -2.316896438598633, + "logps/chosen": -14.563620567321777, + "logps/rejected": -15.855158805847168, + "loss": 1.1342, + "rewards/accuracies": 0.59375, + "rewards/chosen": 0.07377764582633972, + "rewards/margins": 0.02341487631201744, + "rewards/rejected": 0.05036277323961258, + "step": 105 + }, + { + "epoch": 0.2327113062568606, + "grad_norm": 1.5053890943527222, + "learning_rate": 4.999323102948655e-06, + "logits/chosen": -2.3042502403259277, + "logits/rejected": -2.339005947113037, + "logps/chosen": -12.038972854614258, + "logps/rejected": -14.546398162841797, + "loss": 1.043, + "rewards/accuracies": 0.53125, + "rewards/chosen": 0.06436577439308167, + "rewards/margins": 0.012802021577954292, + "rewards/rejected": 0.051563750952482224, + "step": 106 + }, + { + "epoch": 0.2349066959385291, + "grad_norm": 1.228031039237976, + "learning_rate": 4.999078682916774e-06, + "logits/chosen": -2.3684043884277344, + "logits/rejected": -2.3670406341552734, + "logps/chosen": -12.876321792602539, + "logps/rejected": -9.346763610839844, + "loss": 0.9602, + "rewards/accuracies": 0.5, + "rewards/chosen": 0.019609777256846428, + "rewards/margins": -0.001786000095307827, + "rewards/rejected": 0.021395772695541382, + "step": 107 + }, + { + "epoch": 0.23710208562019758, + "grad_norm": 1.644127607345581, + "learning_rate": 4.998796669702378e-06, + "logits/chosen": -2.303502082824707, + "logits/rejected": -2.3413586616516113, + "logps/chosen": -12.201820373535156, + "logps/rejected": -9.633750915527344, + "loss": 0.9549, + "rewards/accuracies": 0.46875, + "rewards/chosen": 0.06634039431810379, + "rewards/margins": 0.019073188304901123, + "rewards/rejected": 0.04726720601320267, + "step": 108 + }, + { + "epoch": 0.23929747530186607, + "grad_norm": 1.7933754920959473, + "learning_rate": 4.99847706754774e-06, + "logits/chosen": -2.438173294067383, + "logits/rejected": -2.274287700653076, + "logps/chosen": -17.797473907470703, + "logps/rejected": -16.017791748046875, + "loss": 1.0189, + "rewards/accuracies": 0.5, + "rewards/chosen": 0.09196966886520386, + "rewards/margins": 0.05703253298997879, + "rewards/rejected": 0.034937139600515366, + "step": 109 + }, + { + "epoch": 0.2414928649835346, + "grad_norm": 1.5571792125701904, + "learning_rate": 4.998119881260576e-06, + "logits/chosen": -2.316495418548584, + "logits/rejected": -2.309934616088867, + "logps/chosen": -12.530914306640625, + "logps/rejected": -9.023895263671875, + "loss": 0.9321, + "rewards/accuracies": 0.65625, + "rewards/chosen": 0.05945971608161926, + "rewards/margins": 0.03105064295232296, + "rewards/rejected": 0.02840907871723175, + "step": 110 + }, + { + "epoch": 0.24368825466520308, + "grad_norm": 1.5516993999481201, + "learning_rate": 4.997725116213974e-06, + "logits/chosen": -2.220071315765381, + "logits/rejected": -2.2800261974334717, + "logps/chosen": -9.25605583190918, + "logps/rejected": -9.267400741577148, + "loss": 0.9622, + "rewards/accuracies": 0.4375, + "rewards/chosen": 0.05536666512489319, + "rewards/margins": 0.012987159192562103, + "rewards/rejected": 0.042379505932331085, + "step": 111 + }, + { + "epoch": 0.24588364434687157, + "grad_norm": 1.7297463417053223, + "learning_rate": 4.997292778346312e-06, + "logits/chosen": -2.2448363304138184, + "logits/rejected": -2.297600269317627, + "logps/chosen": -11.284765243530273, + "logps/rejected": -12.576556205749512, + "loss": 1.0317, + "rewards/accuracies": 0.5625, + "rewards/chosen": 0.03251136094331741, + "rewards/margins": 0.019679736346006393, + "rewards/rejected": 0.012831625528633595, + "step": 112 + }, + { + "epoch": 0.24807903402854006, + "grad_norm": 1.5127956867218018, + "learning_rate": 4.99682287416117e-06, + "logits/chosen": -2.2955853939056396, + "logits/rejected": -2.3179407119750977, + "logps/chosen": -11.177826881408691, + "logps/rejected": -9.537398338317871, + "loss": 0.9662, + "rewards/accuracies": 0.59375, + "rewards/chosen": 0.06957294791936874, + "rewards/margins": 0.022395242005586624, + "rewards/rejected": 0.04717769846320152, + "step": 113 + }, + { + "epoch": 0.2502744237102086, + "grad_norm": 1.649693489074707, + "learning_rate": 4.996315410727229e-06, + "logits/chosen": -2.377718448638916, + "logits/rejected": -2.32913875579834, + "logps/chosen": -9.667226791381836, + "logps/rejected": -12.393421173095703, + "loss": 1.0772, + "rewards/accuracies": 0.46875, + "rewards/chosen": 0.04138043522834778, + "rewards/margins": -0.03883221000432968, + "rewards/rejected": 0.08021265268325806, + "step": 114 + }, + { + "epoch": 0.2524698133918771, + "grad_norm": 1.8851720094680786, + "learning_rate": 4.995770395678171e-06, + "logits/chosen": -2.2822062969207764, + "logits/rejected": -2.3547139167785645, + "logps/chosen": -12.832666397094727, + "logps/rejected": -10.308067321777344, + "loss": 0.9975, + "rewards/accuracies": 0.5625, + "rewards/chosen": 0.03939870372414589, + "rewards/margins": 0.013363707810640335, + "rewards/rejected": 0.026034995913505554, + "step": 115 + }, + { + "epoch": 0.25466520307354557, + "grad_norm": 1.5929930210113525, + "learning_rate": 4.995187837212555e-06, + "logits/chosen": -2.455446720123291, + "logits/rejected": -2.4105300903320312, + "logps/chosen": -16.525413513183594, + "logps/rejected": -11.672962188720703, + "loss": 1.0094, + "rewards/accuracies": 0.5, + "rewards/chosen": 0.031660135835409164, + "rewards/margins": 0.0169740691781044, + "rewards/rejected": 0.014686062932014465, + "step": 116 + }, + { + "epoch": 0.25686059275521406, + "grad_norm": 1.4787267446517944, + "learning_rate": 4.994567744093703e-06, + "logits/chosen": -2.366934061050415, + "logits/rejected": -2.276400566101074, + "logps/chosen": -11.443145751953125, + "logps/rejected": -12.336971282958984, + "loss": 1.0239, + "rewards/accuracies": 0.4375, + "rewards/chosen": 0.024969782680273056, + "rewards/margins": 0.013532733544707298, + "rewards/rejected": 0.011437049135565758, + "step": 117 + }, + { + "epoch": 0.25905598243688255, + "grad_norm": 1.632069706916809, + "learning_rate": 4.993910125649561e-06, + "logits/chosen": -2.3295586109161377, + "logits/rejected": -2.3176655769348145, + "logps/chosen": -15.515420913696289, + "logps/rejected": -12.011621475219727, + "loss": 1.0096, + "rewards/accuracies": 0.59375, + "rewards/chosen": 0.0905236303806305, + "rewards/margins": 0.0291280597448349, + "rewards/rejected": 0.06139557063579559, + "step": 118 + }, + { + "epoch": 0.26125137211855104, + "grad_norm": 1.6845340728759766, + "learning_rate": 4.993214991772563e-06, + "logits/chosen": -2.3398208618164062, + "logits/rejected": -2.3458690643310547, + "logps/chosen": -12.432971954345703, + "logps/rejected": -8.691479682922363, + "loss": 0.9449, + "rewards/accuracies": 0.40625, + "rewards/chosen": 0.07083068788051605, + "rewards/margins": 0.0008336303289979696, + "rewards/rejected": 0.06999707221984863, + "step": 119 + }, + { + "epoch": 0.26344676180021953, + "grad_norm": 1.4854542016983032, + "learning_rate": 4.99248235291948e-06, + "logits/chosen": -2.3151345252990723, + "logits/rejected": -2.3466830253601074, + "logps/chosen": -12.065990447998047, + "logps/rejected": -11.669464111328125, + "loss": 0.9982, + "rewards/accuracies": 0.625, + "rewards/chosen": 0.03543815016746521, + "rewards/margins": 0.05088028684258461, + "rewards/rejected": -0.01544213853776455, + "step": 120 + }, + { + "epoch": 0.265642151481888, + "grad_norm": 1.7356363534927368, + "learning_rate": 4.991712220111265e-06, + "logits/chosen": -2.3189940452575684, + "logits/rejected": -2.3774425983428955, + "logps/chosen": -14.105989456176758, + "logps/rejected": -10.266246795654297, + "loss": 0.9667, + "rewards/accuracies": 0.65625, + "rewards/chosen": 0.09679834544658661, + "rewards/margins": 0.047457098960876465, + "rewards/rejected": 0.049341246485710144, + "step": 121 + }, + { + "epoch": 0.2678375411635565, + "grad_norm": 1.601832628250122, + "learning_rate": 4.990904604932885e-06, + "logits/chosen": -2.347449541091919, + "logits/rejected": -2.352397918701172, + "logps/chosen": -15.789161682128906, + "logps/rejected": -13.354789733886719, + "loss": 1.0227, + "rewards/accuracies": 0.5625, + "rewards/chosen": 0.09421166777610779, + "rewards/margins": 0.055517613887786865, + "rewards/rejected": 0.03869405388832092, + "step": 122 + }, + { + "epoch": 0.270032930845225, + "grad_norm": 1.4014909267425537, + "learning_rate": 4.990059519533145e-06, + "logits/chosen": -2.2689881324768066, + "logits/rejected": -2.283444404602051, + "logps/chosen": -11.951663970947266, + "logps/rejected": -13.521688461303711, + "loss": 1.077, + "rewards/accuracies": 0.46875, + "rewards/chosen": 0.07996637374162674, + "rewards/margins": 0.0017875898629426956, + "rewards/rejected": 0.0781787857413292, + "step": 123 + }, + { + "epoch": 0.2722283205268935, + "grad_norm": 2.307819128036499, + "learning_rate": 4.989176976624511e-06, + "logits/chosen": -2.313141107559204, + "logits/rejected": -2.3250885009765625, + "logps/chosen": -15.183206558227539, + "logps/rejected": -14.606795310974121, + "loss": 1.1025, + "rewards/accuracies": 0.5625, + "rewards/chosen": 0.06334234774112701, + "rewards/margins": 0.03910152241587639, + "rewards/rejected": 0.024240825325250626, + "step": 124 + }, + { + "epoch": 0.27442371020856204, + "grad_norm": 1.768710970878601, + "learning_rate": 4.9882569894829146e-06, + "logits/chosen": -2.3245012760162354, + "logits/rejected": -2.4108376502990723, + "logps/chosen": -11.341304779052734, + "logps/rejected": -13.046493530273438, + "loss": 1.0653, + "rewards/accuracies": 0.375, + "rewards/chosen": 0.04837167263031006, + "rewards/margins": -0.03817473351955414, + "rewards/rejected": 0.0865464061498642, + "step": 125 + }, + { + "epoch": 0.27661909989023054, + "grad_norm": 1.8172316551208496, + "learning_rate": 4.987299571947554e-06, + "logits/chosen": -2.2869110107421875, + "logits/rejected": -2.3253166675567627, + "logps/chosen": -12.174118041992188, + "logps/rejected": -9.856450080871582, + "loss": 0.9751, + "rewards/accuracies": 0.5, + "rewards/chosen": 0.06659601628780365, + "rewards/margins": -0.006406224332749844, + "rewards/rejected": 0.07300224900245667, + "step": 126 + }, + { + "epoch": 0.278814489571899, + "grad_norm": 1.9271728992462158, + "learning_rate": 4.986304738420684e-06, + "logits/chosen": -2.3674209117889404, + "logits/rejected": -2.3308446407318115, + "logps/chosen": -10.563587188720703, + "logps/rejected": -11.918947219848633, + "loss": 0.9901, + "rewards/accuracies": 0.46875, + "rewards/chosen": 0.06925193965435028, + "rewards/margins": -0.0023591145873069763, + "rewards/rejected": 0.07161105424165726, + "step": 127 + }, + { + "epoch": 0.2810098792535675, + "grad_norm": 1.8149971961975098, + "learning_rate": 4.9852725038674035e-06, + "logits/chosen": -2.33549427986145, + "logits/rejected": -2.2321507930755615, + "logps/chosen": -9.894359588623047, + "logps/rejected": -13.391302108764648, + "loss": 1.0184, + "rewards/accuracies": 0.53125, + "rewards/chosen": 0.06098070740699768, + "rewards/margins": 0.007608053274452686, + "rewards/rejected": 0.05337265133857727, + "step": 128 + }, + { + "epoch": 0.283205268935236, + "grad_norm": 1.5623228549957275, + "learning_rate": 4.9842028838154285e-06, + "logits/chosen": -2.355159282684326, + "logits/rejected": -2.4218826293945312, + "logps/chosen": -16.222064971923828, + "logps/rejected": -9.621826171875, + "loss": 0.9459, + "rewards/accuracies": 0.65625, + "rewards/chosen": 0.04742385074496269, + "rewards/margins": 0.04059514030814171, + "rewards/rejected": 0.006828710436820984, + "step": 129 + }, + { + "epoch": 0.2854006586169045, + "grad_norm": 1.62009859085083, + "learning_rate": 4.983095894354858e-06, + "logits/chosen": -2.2709784507751465, + "logits/rejected": -2.30684757232666, + "logps/chosen": -11.814016342163086, + "logps/rejected": -13.255616188049316, + "loss": 1.0494, + "rewards/accuracies": 0.5625, + "rewards/chosen": 0.0587792843580246, + "rewards/margins": 0.020036276429891586, + "rewards/rejected": 0.03874301165342331, + "step": 130 + }, + { + "epoch": 0.287596048298573, + "grad_norm": 1.7169865369796753, + "learning_rate": 4.9819515521379295e-06, + "logits/chosen": -2.3297133445739746, + "logits/rejected": -2.3426923751831055, + "logps/chosen": -19.01955223083496, + "logps/rejected": -13.051504135131836, + "loss": 1.0026, + "rewards/accuracies": 0.65625, + "rewards/chosen": 0.10157382488250732, + "rewards/margins": 0.054719578474760056, + "rewards/rejected": 0.04685423523187637, + "step": 131 + }, + { + "epoch": 0.2897914379802415, + "grad_norm": 2.0506138801574707, + "learning_rate": 4.980769874378775e-06, + "logits/chosen": -2.2926626205444336, + "logits/rejected": -2.304621696472168, + "logps/chosen": -15.2979097366333, + "logps/rejected": -14.070364952087402, + "loss": 1.045, + "rewards/accuracies": 0.59375, + "rewards/chosen": 0.06757767498493195, + "rewards/margins": 0.027771957218647003, + "rewards/rejected": 0.039805714040994644, + "step": 132 + }, + { + "epoch": 0.29198682766191, + "grad_norm": 1.728746771812439, + "learning_rate": 4.979550878853154e-06, + "logits/chosen": -2.335216999053955, + "logits/rejected": -2.3271849155426025, + "logps/chosen": -12.738531112670898, + "logps/rejected": -10.87563705444336, + "loss": 0.9702, + "rewards/accuracies": 0.59375, + "rewards/chosen": 0.09164644777774811, + "rewards/margins": 0.03838293254375458, + "rewards/rejected": 0.05326350778341293, + "step": 133 + }, + { + "epoch": 0.29418221734357847, + "grad_norm": 1.667359709739685, + "learning_rate": 4.978294583898196e-06, + "logits/chosen": -2.3262503147125244, + "logits/rejected": -2.3686673641204834, + "logps/chosen": -11.43099308013916, + "logps/rejected": -7.651641368865967, + "loss": 0.8551, + "rewards/accuracies": 0.6875, + "rewards/chosen": 0.09794393181800842, + "rewards/margins": 0.10966614633798599, + "rewards/rejected": -0.011722217313945293, + "step": 134 + }, + { + "epoch": 0.29637760702524696, + "grad_norm": 1.7627335786819458, + "learning_rate": 4.977001008412113e-06, + "logits/chosen": -2.2389612197875977, + "logits/rejected": -2.309508800506592, + "logps/chosen": -14.33803653717041, + "logps/rejected": -10.298504829406738, + "loss": 1.0181, + "rewards/accuracies": 0.375, + "rewards/chosen": 0.034048259258270264, + "rewards/margins": -0.03645710274577141, + "rewards/rejected": 0.07050535827875137, + "step": 135 + }, + { + "epoch": 0.29857299670691545, + "grad_norm": 1.8202238082885742, + "learning_rate": 4.975670171853926e-06, + "logits/chosen": -2.2767112255096436, + "logits/rejected": -2.3150486946105957, + "logps/chosen": -11.937854766845703, + "logps/rejected": -14.83053207397461, + "loss": 1.0772, + "rewards/accuracies": 0.59375, + "rewards/chosen": 0.07502683997154236, + "rewards/margins": 0.028351005166769028, + "rewards/rejected": 0.04667583480477333, + "step": 136 + }, + { + "epoch": 0.300768386388584, + "grad_norm": 1.7850346565246582, + "learning_rate": 4.974302094243164e-06, + "logits/chosen": -2.2587356567382812, + "logits/rejected": -2.370450496673584, + "logps/chosen": -14.766382217407227, + "logps/rejected": -11.365856170654297, + "loss": 1.0113, + "rewards/accuracies": 0.59375, + "rewards/chosen": 0.024949384853243828, + "rewards/margins": 0.01844519004225731, + "rewards/rejected": 0.00650419294834137, + "step": 137 + }, + { + "epoch": 0.3029637760702525, + "grad_norm": 1.7888866662979126, + "learning_rate": 4.972896796159568e-06, + "logits/chosen": -2.2733049392700195, + "logits/rejected": -2.3269832134246826, + "logps/chosen": -14.468865394592285, + "logps/rejected": -10.19251823425293, + "loss": 0.9104, + "rewards/accuracies": 0.71875, + "rewards/chosen": 0.11349797993898392, + "rewards/margins": 0.13075613975524902, + "rewards/rejected": -0.01725815236568451, + "step": 138 + }, + { + "epoch": 0.305159165751921, + "grad_norm": 1.5196716785430908, + "learning_rate": 4.971454298742779e-06, + "logits/chosen": -2.30745267868042, + "logits/rejected": -2.258906602859497, + "logps/chosen": -13.690924644470215, + "logps/rejected": -10.550579071044922, + "loss": 0.9217, + "rewards/accuracies": 0.65625, + "rewards/chosen": 0.09285424649715424, + "rewards/margins": 0.08185766637325287, + "rewards/rejected": 0.010996590368449688, + "step": 139 + }, + { + "epoch": 0.30735455543358947, + "grad_norm": 1.8636034727096558, + "learning_rate": 4.969974623692023e-06, + "logits/chosen": -2.2885866165161133, + "logits/rejected": -2.3624653816223145, + "logps/chosen": -15.346282005310059, + "logps/rejected": -13.977486610412598, + "loss": 1.0218, + "rewards/accuracies": 0.65625, + "rewards/chosen": 0.06509467214345932, + "rewards/margins": 0.07762052118778229, + "rewards/rejected": -0.012525845319032669, + "step": 140 + }, + { + "epoch": 0.30954994511525796, + "grad_norm": 1.990831971168518, + "learning_rate": 4.968457793265778e-06, + "logits/chosen": -2.2647054195404053, + "logits/rejected": -2.266617774963379, + "logps/chosen": -12.410196304321289, + "logps/rejected": -10.732589721679688, + "loss": 0.9477, + "rewards/accuracies": 0.53125, + "rewards/chosen": 0.07372469455003738, + "rewards/margins": 0.06349782645702362, + "rewards/rejected": 0.010226882994174957, + "step": 141 + }, + { + "epoch": 0.31174533479692645, + "grad_norm": 1.7295087575912476, + "learning_rate": 4.966903830281449e-06, + "logits/chosen": -2.266833782196045, + "logits/rejected": -2.2585251331329346, + "logps/chosen": -12.859150886535645, + "logps/rejected": -10.204259872436523, + "loss": 0.955, + "rewards/accuracies": 0.5, + "rewards/chosen": 0.05344117432832718, + "rewards/margins": 0.021849703043699265, + "rewards/rejected": 0.03159147500991821, + "step": 142 + }, + { + "epoch": 0.31394072447859495, + "grad_norm": 2.444378137588501, + "learning_rate": 4.965312758115013e-06, + "logits/chosen": -2.391078233718872, + "logits/rejected": -2.3234777450561523, + "logps/chosen": -10.306347846984863, + "logps/rejected": -10.704697608947754, + "loss": 0.9832, + "rewards/accuracies": 0.5625, + "rewards/chosen": 0.0930076315999031, + "rewards/margins": 0.04472024738788605, + "rewards/rejected": 0.04828737676143646, + "step": 143 + }, + { + "epoch": 0.31613611416026344, + "grad_norm": 1.8083428144454956, + "learning_rate": 4.963684600700679e-06, + "logits/chosen": -2.3120150566101074, + "logits/rejected": -2.2713851928710938, + "logps/chosen": -19.03668785095215, + "logps/rejected": -14.014898300170898, + "loss": 1.0903, + "rewards/accuracies": 0.5, + "rewards/chosen": 0.035560242831707, + "rewards/margins": 0.000476097222417593, + "rewards/rejected": 0.03508414700627327, + "step": 144 + }, + { + "epoch": 0.31833150384193193, + "grad_norm": 1.7758944034576416, + "learning_rate": 4.962019382530521e-06, + "logits/chosen": -2.300203800201416, + "logits/rejected": -2.3651368618011475, + "logps/chosen": -13.401725769042969, + "logps/rejected": -12.91569709777832, + "loss": 0.993, + "rewards/accuracies": 0.75, + "rewards/chosen": 0.08328074961900711, + "rewards/margins": 0.12566398084163666, + "rewards/rejected": -0.04238323122262955, + "step": 145 + }, + { + "epoch": 0.3205268935236004, + "grad_norm": 2.1212151050567627, + "learning_rate": 4.9603171286541085e-06, + "logits/chosen": -2.247659206390381, + "logits/rejected": -2.3920531272888184, + "logps/chosen": -12.737861633300781, + "logps/rejected": -14.069719314575195, + "loss": 1.0722, + "rewards/accuracies": 0.5625, + "rewards/chosen": 0.055058207362890244, + "rewards/margins": 0.011270837858319283, + "rewards/rejected": 0.04378737136721611, + "step": 146 + }, + { + "epoch": 0.3227222832052689, + "grad_norm": 2.6743414402008057, + "learning_rate": 4.958577864678137e-06, + "logits/chosen": -2.3820812702178955, + "logits/rejected": -2.429454803466797, + "logps/chosen": -10.022626876831055, + "logps/rejected": -10.506009101867676, + "loss": 0.986, + "rewards/accuracies": 0.65625, + "rewards/chosen": 0.07218223810195923, + "rewards/margins": 0.053717922419309616, + "rewards/rejected": 0.018464315682649612, + "step": 147 + }, + { + "epoch": 0.32491767288693746, + "grad_norm": 2.1124470233917236, + "learning_rate": 4.956801616766033e-06, + "logits/chosen": -2.208386182785034, + "logits/rejected": -2.425445556640625, + "logps/chosen": -17.890907287597656, + "logps/rejected": -15.792840003967285, + "loss": 1.051, + "rewards/accuracies": 0.65625, + "rewards/chosen": 0.08094897866249084, + "rewards/margins": 0.1441258043050766, + "rewards/rejected": -0.06317683309316635, + "step": 148 + }, + { + "epoch": 0.32711306256860595, + "grad_norm": 1.9295481443405151, + "learning_rate": 4.9549884116375714e-06, + "logits/chosen": -2.2656877040863037, + "logits/rejected": -2.328075885772705, + "logps/chosen": -13.346929550170898, + "logps/rejected": -10.23320484161377, + "loss": 0.9307, + "rewards/accuracies": 0.71875, + "rewards/chosen": 0.07452508807182312, + "rewards/margins": 0.08161447942256927, + "rewards/rejected": -0.007089395076036453, + "step": 149 + }, + { + "epoch": 0.32930845225027444, + "grad_norm": 2.049334764480591, + "learning_rate": 4.953138276568462e-06, + "logits/chosen": -2.3590216636657715, + "logits/rejected": -2.3576831817626953, + "logps/chosen": -11.798664093017578, + "logps/rejected": -9.424067497253418, + "loss": 0.9288, + "rewards/accuracies": 0.59375, + "rewards/chosen": 0.07175838202238083, + "rewards/margins": 0.10401518642902374, + "rewards/rejected": -0.032256804406642914, + "step": 150 + }, + { + "epoch": 0.33150384193194293, + "grad_norm": 2.2335071563720703, + "learning_rate": 4.951251239389949e-06, + "logits/chosen": -2.2458577156066895, + "logits/rejected": -2.3633508682250977, + "logps/chosen": -12.633405685424805, + "logps/rejected": -10.60333251953125, + "loss": 0.9717, + "rewards/accuracies": 0.59375, + "rewards/chosen": 0.03975841775536537, + "rewards/margins": 0.08428110182285309, + "rewards/rejected": -0.04452267661690712, + "step": 151 + }, + { + "epoch": 0.3336992316136114, + "grad_norm": 1.876927137374878, + "learning_rate": 4.949327328488385e-06, + "logits/chosen": -2.308839797973633, + "logits/rejected": -2.2872886657714844, + "logps/chosen": -15.553834915161133, + "logps/rejected": -10.210927963256836, + "loss": 0.9448, + "rewards/accuracies": 0.71875, + "rewards/chosen": 0.05748458206653595, + "rewards/margins": 0.07385220378637314, + "rewards/rejected": -0.01636761613190174, + "step": 152 + }, + { + "epoch": 0.3358946212952799, + "grad_norm": 1.784837007522583, + "learning_rate": 4.9473665728048115e-06, + "logits/chosen": -2.364820957183838, + "logits/rejected": -2.425748348236084, + "logps/chosen": -18.728487014770508, + "logps/rejected": -13.96826171875, + "loss": 1.0259, + "rewards/accuracies": 0.71875, + "rewards/chosen": 0.07712395489215851, + "rewards/margins": 0.1511487364768982, + "rewards/rejected": -0.07402478903532028, + "step": 153 + }, + { + "epoch": 0.3380900109769484, + "grad_norm": 2.1200637817382812, + "learning_rate": 4.9453690018345144e-06, + "logits/chosen": -2.302483558654785, + "logits/rejected": -2.368330717086792, + "logps/chosen": -12.537941932678223, + "logps/rejected": -10.460428237915039, + "loss": 0.9672, + "rewards/accuracies": 0.59375, + "rewards/chosen": 0.044165655970573425, + "rewards/margins": 0.0792531818151474, + "rewards/rejected": -0.035087522119283676, + "step": 154 + }, + { + "epoch": 0.3402854006586169, + "grad_norm": 1.7176671028137207, + "learning_rate": 4.943334645626589e-06, + "logits/chosen": -2.2893455028533936, + "logits/rejected": -2.327376365661621, + "logps/chosen": -14.962259292602539, + "logps/rejected": -13.632776260375977, + "loss": 1.0493, + "rewards/accuracies": 0.65625, + "rewards/chosen": 0.08821810781955719, + "rewards/margins": 0.08609063923358917, + "rewards/rejected": 0.002127467654645443, + "step": 155 + }, + { + "epoch": 0.3424807903402854, + "grad_norm": 1.8398821353912354, + "learning_rate": 4.941263534783482e-06, + "logits/chosen": -2.328542709350586, + "logits/rejected": -2.3318638801574707, + "logps/chosen": -16.296972274780273, + "logps/rejected": -10.691251754760742, + "loss": 0.9273, + "rewards/accuracies": 0.71875, + "rewards/chosen": 0.09197843074798584, + "rewards/margins": 0.14689640700817108, + "rewards/rejected": -0.05491796135902405, + "step": 156 + }, + { + "epoch": 0.3446761800219539, + "grad_norm": 1.9241065979003906, + "learning_rate": 4.939155700460536e-06, + "logits/chosen": -2.3499631881713867, + "logits/rejected": -2.3470518589019775, + "logps/chosen": -12.203739166259766, + "logps/rejected": -10.567827224731445, + "loss": 0.9536, + "rewards/accuracies": 0.6875, + "rewards/chosen": 0.09050009399652481, + "rewards/margins": 0.05735006183385849, + "rewards/rejected": 0.03315002843737602, + "step": 157 + }, + { + "epoch": 0.3468715697036224, + "grad_norm": 2.576176166534424, + "learning_rate": 4.937011174365515e-06, + "logits/chosen": -2.324387550354004, + "logits/rejected": -2.363718032836914, + "logps/chosen": -17.173086166381836, + "logps/rejected": -11.478891372680664, + "loss": 0.9861, + "rewards/accuracies": 0.5, + "rewards/chosen": 0.05878392606973648, + "rewards/margins": 0.060986313968896866, + "rewards/rejected": -0.002202393487095833, + "step": 158 + }, + { + "epoch": 0.34906695938529086, + "grad_norm": 2.130946159362793, + "learning_rate": 4.934829988758131e-06, + "logits/chosen": -2.356431007385254, + "logits/rejected": -2.423962354660034, + "logps/chosen": -12.604738235473633, + "logps/rejected": -10.564749717712402, + "loss": 0.9622, + "rewards/accuracies": 0.625, + "rewards/chosen": 0.047928761690855026, + "rewards/margins": 0.08084459602832794, + "rewards/rejected": -0.032915834337472916, + "step": 159 + }, + { + "epoch": 0.3512623490669594, + "grad_norm": 1.962509274482727, + "learning_rate": 4.93261217644956e-06, + "logits/chosen": -2.357512950897217, + "logits/rejected": -2.3551905155181885, + "logps/chosen": -12.000711441040039, + "logps/rejected": -11.452027320861816, + "loss": 1.0009, + "rewards/accuracies": 0.625, + "rewards/chosen": 0.0928552895784378, + "rewards/margins": 0.09134387969970703, + "rewards/rejected": 0.001511402428150177, + "step": 160 + }, + { + "epoch": 0.3534577387486279, + "grad_norm": 2.2321617603302, + "learning_rate": 4.930357770801946e-06, + "logits/chosen": -2.3161845207214355, + "logits/rejected": -2.3024725914001465, + "logps/chosen": -14.647942543029785, + "logps/rejected": -11.785874366760254, + "loss": 0.9646, + "rewards/accuracies": 0.65625, + "rewards/chosen": 0.057551611214876175, + "rewards/margins": 0.09231199324131012, + "rewards/rejected": -0.034760378301143646, + "step": 161 + }, + { + "epoch": 0.3556531284302964, + "grad_norm": 2.041982889175415, + "learning_rate": 4.928066805727901e-06, + "logits/chosen": -2.3434221744537354, + "logits/rejected": -2.3381900787353516, + "logps/chosen": -17.95254135131836, + "logps/rejected": -12.574861526489258, + "loss": 0.9744, + "rewards/accuracies": 0.5625, + "rewards/chosen": 0.0844038724899292, + "rewards/margins": 0.11732687801122665, + "rewards/rejected": -0.032922983169555664, + "step": 162 + }, + { + "epoch": 0.3578485181119649, + "grad_norm": 1.9384185075759888, + "learning_rate": 4.925739315689991e-06, + "logits/chosen": -2.372062921524048, + "logits/rejected": -2.296802282333374, + "logps/chosen": -12.097003936767578, + "logps/rejected": -15.5323486328125, + "loss": 1.0853, + "rewards/accuracies": 0.65625, + "rewards/chosen": 0.03156058490276337, + "rewards/margins": 0.09192690998315811, + "rewards/rejected": -0.06036633625626564, + "step": 163 + }, + { + "epoch": 0.3600439077936334, + "grad_norm": 2.985081911087036, + "learning_rate": 4.923375335700224e-06, + "logits/chosen": -2.35788631439209, + "logits/rejected": -2.3205502033233643, + "logps/chosen": -15.22624397277832, + "logps/rejected": -14.541189193725586, + "loss": 1.0442, + "rewards/accuracies": 0.59375, + "rewards/chosen": 0.035873882472515106, + "rewards/margins": 0.061657097190618515, + "rewards/rejected": -0.02578321285545826, + "step": 164 + }, + { + "epoch": 0.36223929747530187, + "grad_norm": 2.4644775390625, + "learning_rate": 4.9209749013195155e-06, + "logits/chosen": -2.4986884593963623, + "logits/rejected": -2.368385076522827, + "logps/chosen": -13.233194351196289, + "logps/rejected": -12.936639785766602, + "loss": 0.9809, + "rewards/accuracies": 0.625, + "rewards/chosen": 0.047733619809150696, + "rewards/margins": 0.08755230903625488, + "rewards/rejected": -0.03981868550181389, + "step": 165 + }, + { + "epoch": 0.36443468715697036, + "grad_norm": 2.0228378772735596, + "learning_rate": 4.91853804865716e-06, + "logits/chosen": -2.2875237464904785, + "logits/rejected": -2.245499849319458, + "logps/chosen": -13.337098121643066, + "logps/rejected": -11.761133193969727, + "loss": 0.9976, + "rewards/accuracies": 0.53125, + "rewards/chosen": 0.07397015392780304, + "rewards/margins": 0.016120310872793198, + "rewards/rejected": 0.05784984305500984, + "step": 166 + }, + { + "epoch": 0.36663007683863885, + "grad_norm": 2.2122652530670166, + "learning_rate": 4.916064814370287e-06, + "logits/chosen": -2.361201763153076, + "logits/rejected": -2.31585431098938, + "logps/chosen": -21.456907272338867, + "logps/rejected": -13.594123840332031, + "loss": 1.0081, + "rewards/accuracies": 0.59375, + "rewards/chosen": 0.09469065815210342, + "rewards/margins": 0.11242985725402832, + "rewards/rejected": -0.017739199101924896, + "step": 167 + }, + { + "epoch": 0.36882546652030734, + "grad_norm": 2.4223556518554688, + "learning_rate": 4.913555235663306e-06, + "logits/chosen": -2.3535289764404297, + "logits/rejected": -2.330043315887451, + "logps/chosen": -12.632347106933594, + "logps/rejected": -11.649641036987305, + "loss": 0.9651, + "rewards/accuracies": 0.5625, + "rewards/chosen": 0.022385867312550545, + "rewards/margins": 0.11026401072740555, + "rewards/rejected": -0.08787814527750015, + "step": 168 + }, + { + "epoch": 0.37102085620197583, + "grad_norm": 2.7966530323028564, + "learning_rate": 4.911009350287348e-06, + "logits/chosen": -2.3594818115234375, + "logits/rejected": -2.4108834266662598, + "logps/chosen": -14.320435523986816, + "logps/rejected": -8.055628776550293, + "loss": 0.8865, + "rewards/accuracies": 0.59375, + "rewards/chosen": 0.024796217679977417, + "rewards/margins": 0.1393408477306366, + "rewards/rejected": -0.11454464495182037, + "step": 169 + }, + { + "epoch": 0.3732162458836443, + "grad_norm": 3.0437235832214355, + "learning_rate": 4.908427196539701e-06, + "logits/chosen": -2.3775222301483154, + "logits/rejected": -2.3491196632385254, + "logps/chosen": -12.23366928100586, + "logps/rejected": -11.17028522491455, + "loss": 0.9193, + "rewards/accuracies": 0.78125, + "rewards/chosen": 0.0801595002412796, + "rewards/margins": 0.1766997128725052, + "rewards/rejected": -0.09654020518064499, + "step": 170 + }, + { + "epoch": 0.3754116355653128, + "grad_norm": 1.9359002113342285, + "learning_rate": 4.905808813263231e-06, + "logits/chosen": -2.2803163528442383, + "logits/rejected": -2.38509202003479, + "logps/chosen": -14.892003059387207, + "logps/rejected": -9.033920288085938, + "loss": 0.8289, + "rewards/accuracies": 0.78125, + "rewards/chosen": 0.14654552936553955, + "rewards/margins": 0.2833712100982666, + "rewards/rejected": -0.13682566583156586, + "step": 171 + }, + { + "epoch": 0.37760702524698136, + "grad_norm": 2.1269936561584473, + "learning_rate": 4.903154239845798e-06, + "logits/chosen": -2.3570709228515625, + "logits/rejected": -2.3689980506896973, + "logps/chosen": -17.01831817626953, + "logps/rejected": -10.841741561889648, + "loss": 0.9478, + "rewards/accuracies": 0.78125, + "rewards/chosen": 0.05472090095281601, + "rewards/margins": 0.12769103050231934, + "rewards/rejected": -0.07297012954950333, + "step": 172 + }, + { + "epoch": 0.37980241492864986, + "grad_norm": 2.4285571575164795, + "learning_rate": 4.900463516219661e-06, + "logits/chosen": -2.378610610961914, + "logits/rejected": -2.311692714691162, + "logps/chosen": -13.103206634521484, + "logps/rejected": -11.916963577270508, + "loss": 1.0449, + "rewards/accuracies": 0.4375, + "rewards/chosen": 0.01136775128543377, + "rewards/margins": 0.06381823122501373, + "rewards/rejected": -0.05245048552751541, + "step": 173 + }, + { + "epoch": 0.38199780461031835, + "grad_norm": 2.457991600036621, + "learning_rate": 4.897736682860885e-06, + "logits/chosen": -2.392366409301758, + "logits/rejected": -2.4145660400390625, + "logps/chosen": -19.734670639038086, + "logps/rejected": -12.913403511047363, + "loss": 1.0055, + "rewards/accuracies": 0.53125, + "rewards/chosen": 0.005804223008453846, + "rewards/margins": 0.050883978605270386, + "rewards/rejected": -0.04507976025342941, + "step": 174 + }, + { + "epoch": 0.38419319429198684, + "grad_norm": 2.509645462036133, + "learning_rate": 4.894973780788722e-06, + "logits/chosen": -2.38358736038208, + "logits/rejected": -2.460597038269043, + "logps/chosen": -15.367487907409668, + "logps/rejected": -12.089338302612305, + "loss": 0.9987, + "rewards/accuracies": 0.6875, + "rewards/chosen": 0.03781091794371605, + "rewards/margins": 0.16976892948150635, + "rewards/rejected": -0.1319580227136612, + "step": 175 + }, + { + "epoch": 0.38638858397365533, + "grad_norm": 3.260399341583252, + "learning_rate": 4.892174851565004e-06, + "logits/chosen": -2.232401132583618, + "logits/rejected": -2.3769664764404297, + "logps/chosen": -18.106616973876953, + "logps/rejected": -16.98265266418457, + "loss": 1.0731, + "rewards/accuracies": 0.6875, + "rewards/chosen": 0.055242907255887985, + "rewards/margins": 0.12538807094097137, + "rewards/rejected": -0.07014517486095428, + "step": 176 + }, + { + "epoch": 0.3885839736553238, + "grad_norm": 2.5841009616851807, + "learning_rate": 4.889339937293508e-06, + "logits/chosen": -2.274055004119873, + "logits/rejected": -2.2682862281799316, + "logps/chosen": -14.28443717956543, + "logps/rejected": -14.35059642791748, + "loss": 0.9869, + "rewards/accuracies": 0.625, + "rewards/chosen": 0.07298749685287476, + "rewards/margins": 0.15667203068733215, + "rewards/rejected": -0.0836845338344574, + "step": 177 + }, + { + "epoch": 0.3907793633369923, + "grad_norm": 2.899467945098877, + "learning_rate": 4.88646908061933e-06, + "logits/chosen": -2.4936909675598145, + "logits/rejected": -2.4529361724853516, + "logps/chosen": -15.615601539611816, + "logps/rejected": -12.965082168579102, + "loss": 0.9921, + "rewards/accuracies": 0.59375, + "rewards/chosen": -0.025315597653388977, + "rewards/margins": 0.050315871834754944, + "rewards/rejected": -0.07563146948814392, + "step": 178 + }, + { + "epoch": 0.3929747530186608, + "grad_norm": 2.1207797527313232, + "learning_rate": 4.883562324728242e-06, + "logits/chosen": -2.365647792816162, + "logits/rejected": -2.4170095920562744, + "logps/chosen": -15.290233612060547, + "logps/rejected": -11.950180053710938, + "loss": 0.9684, + "rewards/accuracies": 0.65625, + "rewards/chosen": 0.027094591408967972, + "rewards/margins": 0.14753341674804688, + "rewards/rejected": -0.1204388216137886, + "step": 179 + }, + { + "epoch": 0.3951701427003293, + "grad_norm": 2.1420328617095947, + "learning_rate": 4.8806197133460385e-06, + "logits/chosen": -2.2818522453308105, + "logits/rejected": -2.344179153442383, + "logps/chosen": -13.106456756591797, + "logps/rejected": -9.099660873413086, + "loss": 0.8689, + "rewards/accuracies": 0.6875, + "rewards/chosen": 0.1166091114282608, + "rewards/margins": 0.20032554864883423, + "rewards/rejected": -0.08371645212173462, + "step": 180 + }, + { + "epoch": 0.3973655323819978, + "grad_norm": 2.639103651046753, + "learning_rate": 4.8776412907378845e-06, + "logits/chosen": -2.381890296936035, + "logits/rejected": -2.4541854858398438, + "logps/chosen": -18.474212646484375, + "logps/rejected": -13.981678009033203, + "loss": 1.043, + "rewards/accuracies": 0.65625, + "rewards/chosen": 0.037654660642147064, + "rewards/margins": 0.1330977976322174, + "rewards/rejected": -0.09544314444065094, + "step": 181 + }, + { + "epoch": 0.3995609220636663, + "grad_norm": 2.6630892753601074, + "learning_rate": 4.874627101707644e-06, + "logits/chosen": -2.2812371253967285, + "logits/rejected": -2.385373830795288, + "logps/chosen": -15.278980255126953, + "logps/rejected": -11.584487915039062, + "loss": 0.9284, + "rewards/accuracies": 0.71875, + "rewards/chosen": 0.06725232303142548, + "rewards/margins": 0.22635605931282043, + "rewards/rejected": -0.15910373628139496, + "step": 182 + }, + { + "epoch": 0.40175631174533477, + "grad_norm": 2.6720497608184814, + "learning_rate": 4.871577191597211e-06, + "logits/chosen": -2.365732192993164, + "logits/rejected": -2.291288137435913, + "logps/chosen": -14.112415313720703, + "logps/rejected": -9.701397895812988, + "loss": 0.8559, + "rewards/accuracies": 0.6875, + "rewards/chosen": 0.04575119540095329, + "rewards/margins": 0.22594983875751495, + "rewards/rejected": -0.18019863963127136, + "step": 183 + }, + { + "epoch": 0.4039517014270033, + "grad_norm": 2.456178903579712, + "learning_rate": 4.868491606285823e-06, + "logits/chosen": -2.3805899620056152, + "logits/rejected": -2.33213472366333, + "logps/chosen": -11.890176773071289, + "logps/rejected": -10.625673294067383, + "loss": 0.9725, + "rewards/accuracies": 0.59375, + "rewards/chosen": 0.023435473442077637, + "rewards/margins": 0.060660067945718765, + "rewards/rejected": -0.03722459450364113, + "step": 184 + }, + { + "epoch": 0.4061470911086718, + "grad_norm": 2.2524774074554443, + "learning_rate": 4.865370392189377e-06, + "logits/chosen": -2.2781782150268555, + "logits/rejected": -2.4160242080688477, + "logps/chosen": -14.2537841796875, + "logps/rejected": -12.460886001586914, + "loss": 0.9863, + "rewards/accuracies": 0.75, + "rewards/chosen": 0.046875663101673126, + "rewards/margins": 0.1632286012172699, + "rewards/rejected": -0.11635293066501617, + "step": 185 + }, + { + "epoch": 0.4083424807903403, + "grad_norm": 2.891005277633667, + "learning_rate": 4.86221359625972e-06, + "logits/chosen": -2.384573459625244, + "logits/rejected": -2.367485523223877, + "logps/chosen": -18.918542861938477, + "logps/rejected": -11.104225158691406, + "loss": 0.8836, + "rewards/accuracies": 0.78125, + "rewards/chosen": 0.013368407264351845, + "rewards/margins": 0.2821967601776123, + "rewards/rejected": -0.2688283622264862, + "step": 186 + }, + { + "epoch": 0.4105378704720088, + "grad_norm": 2.5016818046569824, + "learning_rate": 4.859021265983959e-06, + "logits/chosen": -2.316922187805176, + "logits/rejected": -2.4440698623657227, + "logps/chosen": -16.58301544189453, + "logps/rejected": -12.081414222717285, + "loss": 0.9647, + "rewards/accuracies": 0.6875, + "rewards/chosen": -0.0013476097956299782, + "rewards/margins": 0.21520553529262543, + "rewards/rejected": -0.21655313670635223, + "step": 187 + }, + { + "epoch": 0.4127332601536773, + "grad_norm": 3.1846582889556885, + "learning_rate": 4.855793449383731e-06, + "logits/chosen": -2.359532356262207, + "logits/rejected": -2.2862062454223633, + "logps/chosen": -13.504922866821289, + "logps/rejected": -11.573217391967773, + "loss": 0.9578, + "rewards/accuracies": 0.5625, + "rewards/chosen": -0.00021760258823633194, + "rewards/margins": 0.11105790734291077, + "rewards/rejected": -0.11127550899982452, + "step": 188 + }, + { + "epoch": 0.4149286498353458, + "grad_norm": 3.166029930114746, + "learning_rate": 4.852530195014489e-06, + "logits/chosen": -2.372904062271118, + "logits/rejected": -2.4071719646453857, + "logps/chosen": -12.410667419433594, + "logps/rejected": -11.904781341552734, + "loss": 0.9805, + "rewards/accuracies": 0.59375, + "rewards/chosen": -0.017314447090029716, + "rewards/margins": 0.07836966216564178, + "rewards/rejected": -0.09568411111831665, + "step": 189 + }, + { + "epoch": 0.41712403951701427, + "grad_norm": 3.263563394546509, + "learning_rate": 4.849231551964771e-06, + "logits/chosen": -2.2438108921051025, + "logits/rejected": -2.3375155925750732, + "logps/chosen": -15.391338348388672, + "logps/rejected": -11.07363510131836, + "loss": 0.9475, + "rewards/accuracies": 0.59375, + "rewards/chosen": 0.02428864873945713, + "rewards/margins": 0.20056959986686707, + "rewards/rejected": -0.17628096044063568, + "step": 190 + }, + { + "epoch": 0.41931942919868276, + "grad_norm": 2.4130396842956543, + "learning_rate": 4.84589756985546e-06, + "logits/chosen": -2.3162131309509277, + "logits/rejected": -2.284975290298462, + "logps/chosen": -16.690502166748047, + "logps/rejected": -13.928424835205078, + "loss": 0.9419, + "rewards/accuracies": 0.71875, + "rewards/chosen": 0.08059751987457275, + "rewards/margins": 0.2925419807434082, + "rewards/rejected": -0.21194449067115784, + "step": 191 + }, + { + "epoch": 0.42151481888035125, + "grad_norm": 4.085582733154297, + "learning_rate": 4.8425282988390376e-06, + "logits/chosen": -2.3132638931274414, + "logits/rejected": -2.3993396759033203, + "logps/chosen": -17.231279373168945, + "logps/rejected": -15.828384399414062, + "loss": 1.1129, + "rewards/accuracies": 0.5625, + "rewards/chosen": -0.004977023229002953, + "rewards/margins": 0.10474638640880585, + "rewards/rejected": -0.10972341150045395, + "step": 192 + }, + { + "epoch": 0.42371020856201974, + "grad_norm": 2.4137845039367676, + "learning_rate": 4.839123789598829e-06, + "logits/chosen": -2.38979434967041, + "logits/rejected": -2.422316074371338, + "logps/chosen": -15.439363479614258, + "logps/rejected": -13.799774169921875, + "loss": 0.9543, + "rewards/accuracies": 0.65625, + "rewards/chosen": -0.01842639595270157, + "rewards/margins": 0.24182362854480743, + "rewards/rejected": -0.2602500319480896, + "step": 193 + }, + { + "epoch": 0.42590559824368823, + "grad_norm": 2.383225202560425, + "learning_rate": 4.835684093348244e-06, + "logits/chosen": -2.4397788047790527, + "logits/rejected": -2.3394742012023926, + "logps/chosen": -14.448860168457031, + "logps/rejected": -10.81193733215332, + "loss": 0.8818, + "rewards/accuracies": 0.71875, + "rewards/chosen": 0.053237978368997574, + "rewards/margins": 0.23682299256324768, + "rewards/rejected": -0.1835850328207016, + "step": 194 + }, + { + "epoch": 0.4281009879253567, + "grad_norm": 2.947072744369507, + "learning_rate": 4.832209261830002e-06, + "logits/chosen": -2.294675827026367, + "logits/rejected": -2.3386130332946777, + "logps/chosen": -14.494888305664062, + "logps/rejected": -11.394132614135742, + "loss": 0.9588, + "rewards/accuracies": 0.59375, + "rewards/chosen": 0.025634365156292915, + "rewards/margins": 0.13824237883090973, + "rewards/rejected": -0.11260801553726196, + "step": 195 + }, + { + "epoch": 0.43029637760702527, + "grad_norm": 2.2794811725616455, + "learning_rate": 4.828699347315357e-06, + "logits/chosen": -2.387575626373291, + "logits/rejected": -2.407410144805908, + "logps/chosen": -16.202159881591797, + "logps/rejected": -10.102056503295898, + "loss": 0.9207, + "rewards/accuracies": 0.6875, + "rewards/chosen": 0.045964036136865616, + "rewards/margins": 0.18496856093406677, + "rewards/rejected": -0.13900452852249146, + "step": 196 + }, + { + "epoch": 0.43249176728869376, + "grad_norm": 2.803621292114258, + "learning_rate": 4.825154402603308e-06, + "logits/chosen": -2.3108205795288086, + "logits/rejected": -2.353505849838257, + "logps/chosen": -17.8997745513916, + "logps/rejected": -12.011569023132324, + "loss": 0.9036, + "rewards/accuracies": 0.75, + "rewards/chosen": 0.06393789499998093, + "rewards/margins": 0.2662951350212097, + "rewards/rejected": -0.20235726237297058, + "step": 197 + }, + { + "epoch": 0.43468715697036225, + "grad_norm": 2.4866416454315186, + "learning_rate": 4.821574481019811e-06, + "logits/chosen": -2.3187174797058105, + "logits/rejected": -2.2712206840515137, + "logps/chosen": -13.027259826660156, + "logps/rejected": -10.942176818847656, + "loss": 0.9522, + "rewards/accuracies": 0.65625, + "rewards/chosen": 0.003394428174942732, + "rewards/margins": 0.1506178379058838, + "rewards/rejected": -0.14722341299057007, + "step": 198 + }, + { + "epoch": 0.43688254665203075, + "grad_norm": 2.9680466651916504, + "learning_rate": 4.817959636416969e-06, + "logits/chosen": -2.3756744861602783, + "logits/rejected": -2.3005049228668213, + "logps/chosen": -11.833829879760742, + "logps/rejected": -11.094474792480469, + "loss": 0.9724, + "rewards/accuracies": 0.59375, + "rewards/chosen": 0.015104535967111588, + "rewards/margins": 0.09097721427679062, + "rewards/rejected": -0.07587268203496933, + "step": 199 + }, + { + "epoch": 0.43907793633369924, + "grad_norm": 2.566525936126709, + "learning_rate": 4.814309923172227e-06, + "logits/chosen": -2.3382346630096436, + "logits/rejected": -2.504864454269409, + "logps/chosen": -15.685405731201172, + "logps/rejected": -11.08444595336914, + "loss": 0.874, + "rewards/accuracies": 0.8125, + "rewards/chosen": 0.05831221491098404, + "rewards/margins": 0.4754192531108856, + "rewards/rejected": -0.4171070158481598, + "step": 200 + }, + { + "epoch": 0.44127332601536773, + "grad_norm": 3.3313262462615967, + "learning_rate": 4.81062539618755e-06, + "logits/chosen": -2.3377225399017334, + "logits/rejected": -2.3022823333740234, + "logps/chosen": -14.626267433166504, + "logps/rejected": -13.372673034667969, + "loss": 0.9706, + "rewards/accuracies": 0.625, + "rewards/chosen": 0.029092811048030853, + "rewards/margins": 0.19668377935886383, + "rewards/rejected": -0.16759096086025238, + "step": 201 + }, + { + "epoch": 0.4434687156970362, + "grad_norm": 2.3386690616607666, + "learning_rate": 4.806906110888606e-06, + "logits/chosen": -2.3637309074401855, + "logits/rejected": -2.3318238258361816, + "logps/chosen": -16.152206420898438, + "logps/rejected": -13.979665756225586, + "loss": 0.9951, + "rewards/accuracies": 0.6875, + "rewards/chosen": 0.06894893944263458, + "rewards/margins": 0.15928399562835693, + "rewards/rejected": -0.09033505618572235, + "step": 202 + }, + { + "epoch": 0.4456641053787047, + "grad_norm": 3.0604872703552246, + "learning_rate": 4.80315212322392e-06, + "logits/chosen": -2.3710412979125977, + "logits/rejected": -2.3815903663635254, + "logps/chosen": -14.69096565246582, + "logps/rejected": -11.106029510498047, + "loss": 0.9239, + "rewards/accuracies": 0.6875, + "rewards/chosen": -0.012477545998990536, + "rewards/margins": 0.23910902440547943, + "rewards/rejected": -0.25158655643463135, + "step": 203 + }, + { + "epoch": 0.4478594950603732, + "grad_norm": 4.03364896774292, + "learning_rate": 4.799363489664039e-06, + "logits/chosen": -2.3909597396850586, + "logits/rejected": -2.368131160736084, + "logps/chosen": -13.064891815185547, + "logps/rejected": -13.199899673461914, + "loss": 1.0319, + "rewards/accuracies": 0.5625, + "rewards/chosen": 0.029866386204957962, + "rewards/margins": 0.08630535006523132, + "rewards/rejected": -0.05643896758556366, + "step": 204 + }, + { + "epoch": 0.4500548847420417, + "grad_norm": 3.311805248260498, + "learning_rate": 4.7955402672006855e-06, + "logits/chosen": -2.2783560752868652, + "logits/rejected": -2.4314398765563965, + "logps/chosen": -18.015016555786133, + "logps/rejected": -12.451406478881836, + "loss": 0.9464, + "rewards/accuracies": 0.71875, + "rewards/chosen": -0.03410699963569641, + "rewards/margins": 0.30977538228034973, + "rewards/rejected": -0.34388238191604614, + "step": 205 + }, + { + "epoch": 0.4522502744237102, + "grad_norm": 2.7831180095672607, + "learning_rate": 4.7916825133458925e-06, + "logits/chosen": -2.4114174842834473, + "logits/rejected": -2.263611078262329, + "logps/chosen": -15.84518814086914, + "logps/rejected": -12.624719619750977, + "loss": 0.9274, + "rewards/accuracies": 0.65625, + "rewards/chosen": 0.025281542912125587, + "rewards/margins": 0.3231895864009857, + "rewards/rejected": -0.2979080379009247, + "step": 206 + }, + { + "epoch": 0.4544456641053787, + "grad_norm": 3.6785452365875244, + "learning_rate": 4.787790286131145e-06, + "logits/chosen": -2.3511180877685547, + "logits/rejected": -2.340761661529541, + "logps/chosen": -14.801597595214844, + "logps/rejected": -13.564799308776855, + "loss": 0.9737, + "rewards/accuracies": 0.65625, + "rewards/chosen": -0.03549131006002426, + "rewards/margins": 0.19460512697696686, + "rewards/rejected": -0.23009642958641052, + "step": 207 + }, + { + "epoch": 0.4566410537870472, + "grad_norm": 3.273648977279663, + "learning_rate": 4.783863644106502e-06, + "logits/chosen": -2.340733051300049, + "logits/rejected": -2.3513708114624023, + "logps/chosen": -14.800409317016602, + "logps/rejected": -13.056378364562988, + "loss": 1.0409, + "rewards/accuracies": 0.59375, + "rewards/chosen": -0.04681443050503731, + "rewards/margins": 0.048753753304481506, + "rewards/rejected": -0.09556818753480911, + "step": 208 + }, + { + "epoch": 0.4588364434687157, + "grad_norm": 3.515641450881958, + "learning_rate": 4.779902646339722e-06, + "logits/chosen": -2.385969638824463, + "logits/rejected": -2.296189308166504, + "logps/chosen": -13.113666534423828, + "logps/rejected": -10.047690391540527, + "loss": 0.945, + "rewards/accuracies": 0.59375, + "rewards/chosen": 0.012289118953049183, + "rewards/margins": 0.09538790583610535, + "rewards/rejected": -0.08309878408908844, + "step": 209 + }, + { + "epoch": 0.4610318331503842, + "grad_norm": 4.205086708068848, + "learning_rate": 4.775907352415367e-06, + "logits/chosen": -2.3117594718933105, + "logits/rejected": -2.3786885738372803, + "logps/chosen": -13.62367057800293, + "logps/rejected": -10.471162796020508, + "loss": 0.926, + "rewards/accuracies": 0.6875, + "rewards/chosen": -0.009793296456336975, + "rewards/margins": 0.14466646313667297, + "rewards/rejected": -0.15445974469184875, + "step": 210 + }, + { + "epoch": 0.4632272228320527, + "grad_norm": 4.027493476867676, + "learning_rate": 4.7718778224339115e-06, + "logits/chosen": -2.3667526245117188, + "logits/rejected": -2.3443357944488525, + "logps/chosen": -14.223626136779785, + "logps/rejected": -17.153335571289062, + "loss": 1.1377, + "rewards/accuracies": 0.625, + "rewards/chosen": -0.012900881469249725, + "rewards/margins": 0.1315092295408249, + "rewards/rejected": -0.1444101184606552, + "step": 211 + }, + { + "epoch": 0.4654226125137212, + "grad_norm": 4.006818771362305, + "learning_rate": 4.767814117010835e-06, + "logits/chosen": -2.4249491691589355, + "logits/rejected": -2.280056953430176, + "logps/chosen": -13.195720672607422, + "logps/rejected": -8.659795761108398, + "loss": 0.8623, + "rewards/accuracies": 0.71875, + "rewards/chosen": -0.005751830525696278, + "rewards/margins": 0.2284858226776123, + "rewards/rejected": -0.2342376559972763, + "step": 212 + }, + { + "epoch": 0.4676180021953897, + "grad_norm": 2.848663568496704, + "learning_rate": 4.763716297275715e-06, + "logits/chosen": -2.271615982055664, + "logits/rejected": -2.3262381553649902, + "logps/chosen": -13.699186325073242, + "logps/rejected": -10.280715942382812, + "loss": 0.8926, + "rewards/accuracies": 0.71875, + "rewards/chosen": 0.0830225721001625, + "rewards/margins": 0.2727515995502472, + "rewards/rejected": -0.18972903490066528, + "step": 213 + }, + { + "epoch": 0.4698133918770582, + "grad_norm": 2.8376808166503906, + "learning_rate": 4.759584424871302e-06, + "logits/chosen": -2.308769941329956, + "logits/rejected": -2.313756227493286, + "logps/chosen": -15.752427101135254, + "logps/rejected": -10.513671875, + "loss": 0.8535, + "rewards/accuracies": 0.75, + "rewards/chosen": 0.17882277071475983, + "rewards/margins": 0.2885429561138153, + "rewards/rejected": -0.10972017049789429, + "step": 214 + }, + { + "epoch": 0.47200878155872666, + "grad_norm": 2.9638876914978027, + "learning_rate": 4.755418561952595e-06, + "logits/chosen": -2.3878135681152344, + "logits/rejected": -2.4203245639801025, + "logps/chosen": -15.204294204711914, + "logps/rejected": -15.910358428955078, + "loss": 1.0671, + "rewards/accuracies": 0.6875, + "rewards/chosen": 0.03764279559254646, + "rewards/margins": 0.19274656474590302, + "rewards/rejected": -0.15510375797748566, + "step": 215 + }, + { + "epoch": 0.47420417124039516, + "grad_norm": 3.096339464187622, + "learning_rate": 4.751218771185906e-06, + "logits/chosen": -2.3016912937164307, + "logits/rejected": -2.4041194915771484, + "logps/chosen": -17.33173179626465, + "logps/rejected": -11.76982307434082, + "loss": 0.9436, + "rewards/accuracies": 0.6875, + "rewards/chosen": 0.051540665328502655, + "rewards/margins": 0.16486641764640808, + "rewards/rejected": -0.11332575976848602, + "step": 216 + }, + { + "epoch": 0.47639956092206365, + "grad_norm": 2.5373213291168213, + "learning_rate": 4.746985115747918e-06, + "logits/chosen": -2.3217809200286865, + "logits/rejected": -2.3968961238861084, + "logps/chosen": -15.51959228515625, + "logps/rejected": -9.341879844665527, + "loss": 0.8685, + "rewards/accuracies": 0.65625, + "rewards/chosen": -0.01756952330470085, + "rewards/margins": 0.3904784321784973, + "rewards/rejected": -0.40804794430732727, + "step": 217 + }, + { + "epoch": 0.47859495060373214, + "grad_norm": 3.3946480751037598, + "learning_rate": 4.742717659324734e-06, + "logits/chosen": -2.3248724937438965, + "logits/rejected": -2.3879940509796143, + "logps/chosen": -16.362497329711914, + "logps/rejected": -11.980215072631836, + "loss": 0.9029, + "rewards/accuracies": 0.78125, + "rewards/chosen": 0.07993088662624359, + "rewards/margins": 0.30833542346954346, + "rewards/rejected": -0.22840453684329987, + "step": 218 + }, + { + "epoch": 0.4807903402854007, + "grad_norm": 3.332834482192993, + "learning_rate": 4.738416466110918e-06, + "logits/chosen": -2.2557213306427, + "logits/rejected": -2.289715051651001, + "logps/chosen": -15.303863525390625, + "logps/rejected": -11.47325325012207, + "loss": 0.974, + "rewards/accuracies": 0.5625, + "rewards/chosen": 0.01493862271308899, + "rewards/margins": 0.1497972160577774, + "rewards/rejected": -0.1348586082458496, + "step": 219 + }, + { + "epoch": 0.4829857299670692, + "grad_norm": 3.7161571979522705, + "learning_rate": 4.734081600808531e-06, + "logits/chosen": -2.410463809967041, + "logits/rejected": -2.3640031814575195, + "logps/chosen": -17.304058074951172, + "logps/rejected": -9.83676528930664, + "loss": 0.8968, + "rewards/accuracies": 0.71875, + "rewards/chosen": 0.09258735924959183, + "rewards/margins": 0.16851232945919037, + "rewards/rejected": -0.07592497020959854, + "step": 220 + }, + { + "epoch": 0.48518111964873767, + "grad_norm": 4.089512348175049, + "learning_rate": 4.729713128626158e-06, + "logits/chosen": -2.3158750534057617, + "logits/rejected": -2.3622100353240967, + "logps/chosen": -15.238099098205566, + "logps/rejected": -12.259599685668945, + "loss": 0.9982, + "rewards/accuracies": 0.59375, + "rewards/chosen": -0.05108126252889633, + "rewards/margins": 0.20846769213676453, + "rewards/rejected": -0.25954896211624146, + "step": 221 + }, + { + "epoch": 0.48737650933040616, + "grad_norm": 3.147043466567993, + "learning_rate": 4.725311115277924e-06, + "logits/chosen": -2.339972972869873, + "logits/rejected": -2.3009352684020996, + "logps/chosen": -19.9134521484375, + "logps/rejected": -19.152158737182617, + "loss": 1.1412, + "rewards/accuracies": 0.6875, + "rewards/chosen": 0.03218366950750351, + "rewards/margins": 0.30710160732269287, + "rewards/rejected": -0.27491796016693115, + "step": 222 + }, + { + "epoch": 0.48957189901207465, + "grad_norm": 3.934567451477051, + "learning_rate": 4.720875626982511e-06, + "logits/chosen": -2.3560261726379395, + "logits/rejected": -2.2281508445739746, + "logps/chosen": -12.923276901245117, + "logps/rejected": -9.938714981079102, + "loss": 0.893, + "rewards/accuracies": 0.65625, + "rewards/chosen": 0.04361802339553833, + "rewards/margins": 0.1802201271057129, + "rewards/rejected": -0.13660210371017456, + "step": 223 + }, + { + "epoch": 0.49176728869374314, + "grad_norm": 3.16060733795166, + "learning_rate": 4.716406730462154e-06, + "logits/chosen": -2.377735137939453, + "logits/rejected": -2.425727128982544, + "logps/chosen": -16.463150024414062, + "logps/rejected": -11.011037826538086, + "loss": 0.936, + "rewards/accuracies": 0.71875, + "rewards/chosen": 0.021465934813022614, + "rewards/margins": 0.27524274587631226, + "rewards/rejected": -0.25377681851387024, + "step": 224 + }, + { + "epoch": 0.49396267837541163, + "grad_norm": 2.8043062686920166, + "learning_rate": 4.711904492941644e-06, + "logits/chosen": -2.3855056762695312, + "logits/rejected": -2.2830467224121094, + "logps/chosen": -11.66223430633545, + "logps/rejected": -9.482926368713379, + "loss": 0.908, + "rewards/accuracies": 0.65625, + "rewards/chosen": 0.007894640788435936, + "rewards/margins": 0.11858849227428436, + "rewards/rejected": -0.11069385707378387, + "step": 225 + }, + { + "epoch": 0.4961580680570801, + "grad_norm": 2.8552234172821045, + "learning_rate": 4.707368982147318e-06, + "logits/chosen": -2.360535144805908, + "logits/rejected": -2.43141508102417, + "logps/chosen": -12.100669860839844, + "logps/rejected": -9.97754955291748, + "loss": 0.9084, + "rewards/accuracies": 0.5, + "rewards/chosen": 0.005531699396669865, + "rewards/margins": 0.20804524421691895, + "rewards/rejected": -0.20251353085041046, + "step": 226 + }, + { + "epoch": 0.4983534577387486, + "grad_norm": 2.832981824874878, + "learning_rate": 4.7028002663060305e-06, + "logits/chosen": -2.2898478507995605, + "logits/rejected": -2.3290865421295166, + "logps/chosen": -13.829479217529297, + "logps/rejected": -9.624303817749023, + "loss": 0.8996, + "rewards/accuracies": 0.5625, + "rewards/chosen": 0.0005602305755019188, + "rewards/margins": 0.21266406774520874, + "rewards/rejected": -0.21210385859012604, + "step": 227 + }, + { + "epoch": 0.5005488474204172, + "grad_norm": 3.762120008468628, + "learning_rate": 4.698198414144136e-06, + "logits/chosen": -2.3622140884399414, + "logits/rejected": -2.3753628730773926, + "logps/chosen": -16.06410789489746, + "logps/rejected": -9.198554992675781, + "loss": 0.8361, + "rewards/accuracies": 0.65625, + "rewards/chosen": -0.011656716465950012, + "rewards/margins": 0.3140515983104706, + "rewards/rejected": -0.3257082998752594, + "step": 228 + }, + { + "epoch": 0.5027442371020856, + "grad_norm": 2.261446237564087, + "learning_rate": 4.693563494886455e-06, + "logits/chosen": -2.3106729984283447, + "logits/rejected": -2.3571410179138184, + "logps/chosen": -15.244176864624023, + "logps/rejected": -12.582070350646973, + "loss": 0.9288, + "rewards/accuracies": 0.6875, + "rewards/chosen": 0.08472658693790436, + "rewards/margins": 0.2612376809120178, + "rewards/rejected": -0.17651110887527466, + "step": 229 + }, + { + "epoch": 0.5049396267837541, + "grad_norm": 4.166963577270508, + "learning_rate": 4.688895578255228e-06, + "logits/chosen": -2.3666067123413086, + "logits/rejected": -2.3383450508117676, + "logps/chosen": -15.07994556427002, + "logps/rejected": -10.779797554016113, + "loss": 0.8831, + "rewards/accuracies": 0.71875, + "rewards/chosen": -0.017674017697572708, + "rewards/margins": 0.330182820558548, + "rewards/rejected": -0.3478568494319916, + "step": 230 + }, + { + "epoch": 0.5071350164654226, + "grad_norm": 3.2102549076080322, + "learning_rate": 4.684194734469067e-06, + "logits/chosen": -2.310776710510254, + "logits/rejected": -2.341068744659424, + "logps/chosen": -16.106998443603516, + "logps/rejected": -12.700186729431152, + "loss": 0.9787, + "rewards/accuracies": 0.5625, + "rewards/chosen": -0.006546609103679657, + "rewards/margins": 0.25047093629837036, + "rewards/rejected": -0.2570175528526306, + "step": 231 + }, + { + "epoch": 0.5093304061470911, + "grad_norm": 3.613203287124634, + "learning_rate": 4.679461034241906e-06, + "logits/chosen": -2.3264687061309814, + "logits/rejected": -2.2634031772613525, + "logps/chosen": -15.22059154510498, + "logps/rejected": -14.103819847106934, + "loss": 1.0352, + "rewards/accuracies": 0.53125, + "rewards/chosen": 0.014797629788517952, + "rewards/margins": 0.1425546407699585, + "rewards/rejected": -0.1277570128440857, + "step": 232 + }, + { + "epoch": 0.5115257958287596, + "grad_norm": 2.7349843978881836, + "learning_rate": 4.674694548781929e-06, + "logits/chosen": -2.39654541015625, + "logits/rejected": -2.4142558574676514, + "logps/chosen": -17.5054931640625, + "logps/rejected": -12.546804428100586, + "loss": 0.9873, + "rewards/accuracies": 0.53125, + "rewards/chosen": 0.007637964561581612, + "rewards/margins": 0.15906468033790588, + "rewards/rejected": -0.15142671763896942, + "step": 233 + }, + { + "epoch": 0.5137211855104281, + "grad_norm": 4.179075241088867, + "learning_rate": 4.669895349790502e-06, + "logits/chosen": -2.4446206092834473, + "logits/rejected": -2.3479771614074707, + "logps/chosen": -15.197080612182617, + "logps/rejected": -10.906547546386719, + "loss": 0.9273, + "rewards/accuracies": 0.65625, + "rewards/chosen": 0.011189254932105541, + "rewards/margins": 0.14327660202980042, + "rewards/rejected": -0.1320873498916626, + "step": 234 + }, + { + "epoch": 0.5159165751920965, + "grad_norm": 2.9067695140838623, + "learning_rate": 4.665063509461098e-06, + "logits/chosen": -2.3557796478271484, + "logits/rejected": -2.271129608154297, + "logps/chosen": -13.00634765625, + "logps/rejected": -11.378231048583984, + "loss": 0.9386, + "rewards/accuracies": 0.65625, + "rewards/chosen": 0.07278364151716232, + "rewards/margins": 0.18158170580863953, + "rewards/rejected": -0.10879804939031601, + "step": 235 + }, + { + "epoch": 0.5181119648737651, + "grad_norm": 3.2804477214813232, + "learning_rate": 4.660199100478201e-06, + "logits/chosen": -2.3077774047851562, + "logits/rejected": -2.364297389984131, + "logps/chosen": -16.013919830322266, + "logps/rejected": -12.880232810974121, + "loss": 0.9551, + "rewards/accuracies": 0.6875, + "rewards/chosen": 0.02138105407357216, + "rewards/margins": 0.24333827197551727, + "rewards/rejected": -0.2219572216272354, + "step": 236 + }, + { + "epoch": 0.5203073545554336, + "grad_norm": 3.5774881839752197, + "learning_rate": 4.655302196016228e-06, + "logits/chosen": -2.3459157943725586, + "logits/rejected": -2.348891019821167, + "logps/chosen": -12.117788314819336, + "logps/rejected": -12.449483871459961, + "loss": 0.9679, + "rewards/accuracies": 0.625, + "rewards/chosen": 0.02035675011575222, + "rewards/margins": 0.2209654599428177, + "rewards/rejected": -0.20060870051383972, + "step": 237 + }, + { + "epoch": 0.5225027442371021, + "grad_norm": 3.279064893722534, + "learning_rate": 4.650372869738415e-06, + "logits/chosen": -2.3156638145446777, + "logits/rejected": -2.3022217750549316, + "logps/chosen": -14.40345573425293, + "logps/rejected": -13.11058235168457, + "loss": 0.9551, + "rewards/accuracies": 0.65625, + "rewards/chosen": 0.05910070985555649, + "rewards/margins": 0.19128769636154175, + "rewards/rejected": -0.13218699395656586, + "step": 238 + }, + { + "epoch": 0.5246981339187706, + "grad_norm": 3.685303211212158, + "learning_rate": 4.645411195795709e-06, + "logits/chosen": -2.257229804992676, + "logits/rejected": -2.2816812992095947, + "logps/chosen": -14.475461959838867, + "logps/rejected": -8.206293106079102, + "loss": 0.8199, + "rewards/accuracies": 0.8125, + "rewards/chosen": 0.013962533324956894, + "rewards/margins": 0.29057538509368896, + "rewards/rejected": -0.27661287784576416, + "step": 239 + }, + { + "epoch": 0.5268935236004391, + "grad_norm": 2.90091609954834, + "learning_rate": 4.640417248825667e-06, + "logits/chosen": -2.3027405738830566, + "logits/rejected": -2.364433765411377, + "logps/chosen": -12.266948699951172, + "logps/rejected": -12.551021575927734, + "loss": 0.9237, + "rewards/accuracies": 0.78125, + "rewards/chosen": 0.10769608616828918, + "rewards/margins": 0.23384669423103333, + "rewards/rejected": -0.12615060806274414, + "step": 240 + }, + { + "epoch": 0.5290889132821076, + "grad_norm": 3.0138938426971436, + "learning_rate": 4.635391103951315e-06, + "logits/chosen": -2.3460533618927, + "logits/rejected": -2.292667865753174, + "logps/chosen": -18.938508987426758, + "logps/rejected": -16.88079261779785, + "loss": 0.9813, + "rewards/accuracies": 0.8125, + "rewards/chosen": 0.07211557030677795, + "rewards/margins": 0.37093961238861084, + "rewards/rejected": -0.2988240718841553, + "step": 241 + }, + { + "epoch": 0.531284302963776, + "grad_norm": 3.351609945297241, + "learning_rate": 4.630332836780029e-06, + "logits/chosen": -2.3035905361175537, + "logits/rejected": -2.323580026626587, + "logps/chosen": -16.21466827392578, + "logps/rejected": -13.248044967651367, + "loss": 0.9519, + "rewards/accuracies": 0.59375, + "rewards/chosen": 0.01800696738064289, + "rewards/margins": 0.2727009952068329, + "rewards/rejected": -0.25469401478767395, + "step": 242 + }, + { + "epoch": 0.5334796926454446, + "grad_norm": 3.892470121383667, + "learning_rate": 4.625242523402395e-06, + "logits/chosen": -2.347764730453491, + "logits/rejected": -2.4785423278808594, + "logps/chosen": -21.10007095336914, + "logps/rejected": -13.007464408874512, + "loss": 0.9527, + "rewards/accuracies": 0.6875, + "rewards/chosen": -0.026116278022527695, + "rewards/margins": 0.23999722301959991, + "rewards/rejected": -0.2661135196685791, + "step": 243 + }, + { + "epoch": 0.535675082327113, + "grad_norm": 4.164974212646484, + "learning_rate": 4.620120240391065e-06, + "logits/chosen": -2.343069553375244, + "logits/rejected": -2.4068026542663574, + "logps/chosen": -14.178921699523926, + "logps/rejected": -9.567798614501953, + "loss": 0.9327, + "rewards/accuracies": 0.625, + "rewards/chosen": -0.03248784318566322, + "rewards/margins": 0.1319328397512436, + "rewards/rejected": -0.1644206941127777, + "step": 244 + }, + { + "epoch": 0.5378704720087816, + "grad_norm": 3.250154972076416, + "learning_rate": 4.614966064799603e-06, + "logits/chosen": -2.3875789642333984, + "logits/rejected": -2.350403308868408, + "logps/chosen": -14.76641845703125, + "logps/rejected": -11.900374412536621, + "loss": 0.8904, + "rewards/accuracies": 0.8125, + "rewards/chosen": 0.07907561212778091, + "rewards/margins": 0.2815009355545044, + "rewards/rejected": -0.20242534577846527, + "step": 245 + }, + { + "epoch": 0.54006586169045, + "grad_norm": 3.4762589931488037, + "learning_rate": 4.609780074161327e-06, + "logits/chosen": -2.3251564502716064, + "logits/rejected": -2.4273359775543213, + "logps/chosen": -15.190896034240723, + "logps/rejected": -10.03155517578125, + "loss": 0.8642, + "rewards/accuracies": 0.6875, + "rewards/chosen": -0.002804434858262539, + "rewards/margins": 0.3568550646305084, + "rewards/rejected": -0.3596595227718353, + "step": 246 + }, + { + "epoch": 0.5422612513721186, + "grad_norm": 2.9021172523498535, + "learning_rate": 4.604562346488144e-06, + "logits/chosen": -2.3237853050231934, + "logits/rejected": -2.3288207054138184, + "logps/chosen": -16.11346435546875, + "logps/rejected": -9.360347747802734, + "loss": 0.7929, + "rewards/accuracies": 0.75, + "rewards/chosen": 0.0847928449511528, + "rewards/margins": 0.44311946630477905, + "rewards/rejected": -0.35832661390304565, + "step": 247 + }, + { + "epoch": 0.544456641053787, + "grad_norm": 5.830170154571533, + "learning_rate": 4.599312960269375e-06, + "logits/chosen": -2.366364002227783, + "logits/rejected": -2.4149107933044434, + "logps/chosen": -17.478118896484375, + "logps/rejected": -13.510588645935059, + "loss": 1.0225, + "rewards/accuracies": 0.59375, + "rewards/chosen": -0.06305301934480667, + "rewards/margins": 0.21186582744121552, + "rewards/rejected": -0.2749188542366028, + "step": 248 + }, + { + "epoch": 0.5466520307354555, + "grad_norm": 3.584066867828369, + "learning_rate": 4.594031994470574e-06, + "logits/chosen": -2.325279474258423, + "logits/rejected": -2.456352710723877, + "logps/chosen": -15.935647964477539, + "logps/rejected": -15.036849975585938, + "loss": 1.0898, + "rewards/accuracies": 0.625, + "rewards/chosen": -0.015486141666769981, + "rewards/margins": 0.21688403189182281, + "rewards/rejected": -0.23237019777297974, + "step": 249 + }, + { + "epoch": 0.5488474204171241, + "grad_norm": 3.663219928741455, + "learning_rate": 4.588719528532342e-06, + "logits/chosen": -2.4341042041778564, + "logits/rejected": -2.4564149379730225, + "logps/chosen": -16.782196044921875, + "logps/rejected": -10.696434020996094, + "loss": 0.8959, + "rewards/accuracies": 0.71875, + "rewards/chosen": 0.04107421636581421, + "rewards/margins": 0.36102810502052307, + "rewards/rejected": -0.31995388865470886, + "step": 250 + }, + { + "epoch": 0.5510428100987925, + "grad_norm": 3.9447195529937744, + "learning_rate": 4.583375642369129e-06, + "logits/chosen": -2.3444766998291016, + "logits/rejected": -2.3383076190948486, + "logps/chosen": -20.937177658081055, + "logps/rejected": -13.723505020141602, + "loss": 1.0103, + "rewards/accuracies": 0.65625, + "rewards/chosen": 0.021991277113556862, + "rewards/margins": 0.34398943185806274, + "rewards/rejected": -0.32199811935424805, + "step": 251 + }, + { + "epoch": 0.5532381997804611, + "grad_norm": 3.336824417114258, + "learning_rate": 4.5780004163680365e-06, + "logits/chosen": -2.304708480834961, + "logits/rejected": -2.3940112590789795, + "logps/chosen": -18.0738525390625, + "logps/rejected": -11.277729034423828, + "loss": 0.8595, + "rewards/accuracies": 0.8125, + "rewards/chosen": 0.0800776332616806, + "rewards/margins": 0.38721710443496704, + "rewards/rejected": -0.30713948607444763, + "step": 252 + }, + { + "epoch": 0.5554335894621295, + "grad_norm": 3.5766406059265137, + "learning_rate": 4.572593931387604e-06, + "logits/chosen": -2.2744007110595703, + "logits/rejected": -2.3453307151794434, + "logps/chosen": -16.072280883789062, + "logps/rejected": -11.154439926147461, + "loss": 0.956, + "rewards/accuracies": 0.53125, + "rewards/chosen": -0.03496705740690231, + "rewards/margins": 0.132221058011055, + "rewards/rejected": -0.16718809306621552, + "step": 253 + }, + { + "epoch": 0.557628979143798, + "grad_norm": 3.511826515197754, + "learning_rate": 4.567156268756594e-06, + "logits/chosen": -2.4133963584899902, + "logits/rejected": -2.383863687515259, + "logps/chosen": -14.50227165222168, + "logps/rejected": -10.799485206604004, + "loss": 0.8542, + "rewards/accuracies": 0.90625, + "rewards/chosen": 0.021454015746712685, + "rewards/margins": 0.36027950048446655, + "rewards/rejected": -0.3388254642486572, + "step": 254 + }, + { + "epoch": 0.5598243688254665, + "grad_norm": 4.181152820587158, + "learning_rate": 4.561687510272767e-06, + "logits/chosen": -2.2828900814056396, + "logits/rejected": -2.370487928390503, + "logps/chosen": -15.107242584228516, + "logps/rejected": -9.730294227600098, + "loss": 0.9018, + "rewards/accuracies": 0.6875, + "rewards/chosen": -0.022096818313002586, + "rewards/margins": 0.29161518812179565, + "rewards/rejected": -0.3137120008468628, + "step": 255 + }, + { + "epoch": 0.562019758507135, + "grad_norm": 4.829276084899902, + "learning_rate": 4.556187738201656e-06, + "logits/chosen": -2.430100917816162, + "logits/rejected": -2.4558093547821045, + "logps/chosen": -14.387591361999512, + "logps/rejected": -11.638925552368164, + "loss": 0.931, + "rewards/accuracies": 0.71875, + "rewards/chosen": -0.004484318196773529, + "rewards/margins": 0.24601556360721588, + "rewards/rejected": -0.2504999041557312, + "step": 256 + }, + { + "epoch": 0.5642151481888035, + "grad_norm": 3.4663209915161133, + "learning_rate": 4.550657035275323e-06, + "logits/chosen": -2.2517337799072266, + "logits/rejected": -2.3557732105255127, + "logps/chosen": -15.434856414794922, + "logps/rejected": -10.484879493713379, + "loss": 0.8677, + "rewards/accuracies": 0.75, + "rewards/chosen": 0.014802820980548859, + "rewards/margins": 0.32399383187294006, + "rewards/rejected": -0.3091909885406494, + "step": 257 + }, + { + "epoch": 0.566410537870472, + "grad_norm": 4.792181491851807, + "learning_rate": 4.54509548469112e-06, + "logits/chosen": -2.375314712524414, + "logits/rejected": -2.382444381713867, + "logps/chosen": -15.463499069213867, + "logps/rejected": -12.133369445800781, + "loss": 0.9093, + "rewards/accuracies": 0.75, + "rewards/chosen": -0.06232909485697746, + "rewards/margins": 0.34396061301231384, + "rewards/rejected": -0.4062896966934204, + "step": 258 + }, + { + "epoch": 0.5686059275521405, + "grad_norm": 3.6746764183044434, + "learning_rate": 4.539503170110431e-06, + "logits/chosen": -2.3717360496520996, + "logits/rejected": -2.4131274223327637, + "logps/chosen": -17.220232009887695, + "logps/rejected": -13.375675201416016, + "loss": 0.9775, + "rewards/accuracies": 0.75, + "rewards/chosen": -0.01915721222758293, + "rewards/margins": 0.328391969203949, + "rewards/rejected": -0.3475492000579834, + "step": 259 + }, + { + "epoch": 0.570801317233809, + "grad_norm": 3.5798373222351074, + "learning_rate": 4.533880175657419e-06, + "logits/chosen": -2.3118844032287598, + "logits/rejected": -2.2956976890563965, + "logps/chosen": -17.010141372680664, + "logps/rejected": -11.295709609985352, + "loss": 0.8533, + "rewards/accuracies": 0.75, + "rewards/chosen": 0.05849292501807213, + "rewards/margins": 0.3737761080265045, + "rewards/rejected": -0.3152831792831421, + "step": 260 + }, + { + "epoch": 0.5729967069154775, + "grad_norm": 3.199009418487549, + "learning_rate": 4.528226585917761e-06, + "logits/chosen": -2.2721354961395264, + "logits/rejected": -2.3081626892089844, + "logps/chosen": -17.70292091369629, + "logps/rejected": -13.543802261352539, + "loss": 0.8967, + "rewards/accuracies": 0.78125, + "rewards/chosen": 0.05505692958831787, + "rewards/margins": 0.4563320279121399, + "rewards/rejected": -0.401275098323822, + "step": 261 + }, + { + "epoch": 0.575192096597146, + "grad_norm": 3.1294350624084473, + "learning_rate": 4.522542485937369e-06, + "logits/chosen": -2.3046679496765137, + "logits/rejected": -2.344747304916382, + "logps/chosen": -15.264352798461914, + "logps/rejected": -13.952584266662598, + "loss": 0.9476, + "rewards/accuracies": 0.8125, + "rewards/chosen": 0.028087537735700607, + "rewards/margins": 0.25781309604644775, + "rewards/rejected": -0.22972553968429565, + "step": 262 + }, + { + "epoch": 0.5773874862788145, + "grad_norm": 2.8935720920562744, + "learning_rate": 4.516827961221118e-06, + "logits/chosen": -2.311985492706299, + "logits/rejected": -2.3693246841430664, + "logps/chosen": -17.099782943725586, + "logps/rejected": -12.623132705688477, + "loss": 0.9125, + "rewards/accuracies": 0.8125, + "rewards/chosen": 0.10635387897491455, + "rewards/margins": 0.6242843270301819, + "rewards/rejected": -0.5179304480552673, + "step": 263 + }, + { + "epoch": 0.579582875960483, + "grad_norm": 3.569817304611206, + "learning_rate": 4.511083097731556e-06, + "logits/chosen": -2.2364590167999268, + "logits/rejected": -2.280050754547119, + "logps/chosen": -16.519317626953125, + "logps/rejected": -10.577292442321777, + "loss": 0.9376, + "rewards/accuracies": 0.625, + "rewards/chosen": 0.02913329005241394, + "rewards/margins": 0.15598690509796143, + "rewards/rejected": -0.12685361504554749, + "step": 264 + }, + { + "epoch": 0.5817782656421515, + "grad_norm": 3.893734931945801, + "learning_rate": 4.50530798188761e-06, + "logits/chosen": -2.3067941665649414, + "logits/rejected": -2.3932323455810547, + "logps/chosen": -13.595678329467773, + "logps/rejected": -13.469207763671875, + "loss": 1.0158, + "rewards/accuracies": 0.5, + "rewards/chosen": 0.011006025597453117, + "rewards/margins": 0.11442138999700546, + "rewards/rejected": -0.1034153550863266, + "step": 265 + }, + { + "epoch": 0.58397365532382, + "grad_norm": 3.6261839866638184, + "learning_rate": 4.49950270056329e-06, + "logits/chosen": -2.437346935272217, + "logits/rejected": -2.344280958175659, + "logps/chosen": -16.588623046875, + "logps/rejected": -12.61369800567627, + "loss": 0.8737, + "rewards/accuracies": 0.90625, + "rewards/chosen": 0.05177614837884903, + "rewards/margins": 0.39123424887657166, + "rewards/rejected": -0.3394581079483032, + "step": 266 + }, + { + "epoch": 0.5861690450054885, + "grad_norm": 3.539958953857422, + "learning_rate": 4.4936673410863794e-06, + "logits/chosen": -2.4072844982147217, + "logits/rejected": -2.4355082511901855, + "logps/chosen": -21.93024444580078, + "logps/rejected": -12.424371719360352, + "loss": 0.9278, + "rewards/accuracies": 0.78125, + "rewards/chosen": 0.030360229313373566, + "rewards/margins": 0.3428768217563629, + "rewards/rejected": -0.31251657009124756, + "step": 267 + }, + { + "epoch": 0.5883644346871569, + "grad_norm": 4.157306671142578, + "learning_rate": 4.48780199123712e-06, + "logits/chosen": -2.3129444122314453, + "logits/rejected": -2.3434619903564453, + "logps/chosen": -18.06573486328125, + "logps/rejected": -12.912382125854492, + "loss": 0.9321, + "rewards/accuracies": 0.6875, + "rewards/chosen": -0.03493726998567581, + "rewards/margins": 0.283815860748291, + "rewards/rejected": -0.3187531530857086, + "step": 268 + }, + { + "epoch": 0.5905598243688255, + "grad_norm": 4.955633163452148, + "learning_rate": 4.481906739246894e-06, + "logits/chosen": -2.4333293437957764, + "logits/rejected": -2.3597402572631836, + "logps/chosen": -17.311777114868164, + "logps/rejected": -14.16006088256836, + "loss": 1.036, + "rewards/accuracies": 0.625, + "rewards/chosen": -0.01734822988510132, + "rewards/margins": 0.1342713087797165, + "rewards/rejected": -0.1516195386648178, + "step": 269 + }, + { + "epoch": 0.5927552140504939, + "grad_norm": 4.0596022605896, + "learning_rate": 4.475981673796899e-06, + "logits/chosen": -2.4821219444274902, + "logits/rejected": -2.3936054706573486, + "logps/chosen": -15.369222640991211, + "logps/rejected": -11.958950996398926, + "loss": 0.9301, + "rewards/accuracies": 0.6875, + "rewards/chosen": 0.05403044819831848, + "rewards/margins": 0.20187188684940338, + "rewards/rejected": -0.1478414386510849, + "step": 270 + }, + { + "epoch": 0.5949506037321625, + "grad_norm": 3.270981788635254, + "learning_rate": 4.470026884016805e-06, + "logits/chosen": -2.354869842529297, + "logits/rejected": -2.3129818439483643, + "logps/chosen": -13.78475570678711, + "logps/rejected": -11.741534233093262, + "loss": 0.9557, + "rewards/accuracies": 0.75, + "rewards/chosen": 0.0010744757018983364, + "rewards/margins": 0.22390666604042053, + "rewards/rejected": -0.22283217310905457, + "step": 271 + }, + { + "epoch": 0.5971459934138309, + "grad_norm": 4.795289993286133, + "learning_rate": 4.464042459483425e-06, + "logits/chosen": -2.3651349544525146, + "logits/rejected": -2.383972644805908, + "logps/chosen": -16.043739318847656, + "logps/rejected": -12.472469329833984, + "loss": 0.9447, + "rewards/accuracies": 0.71875, + "rewards/chosen": 0.0016098134219646454, + "rewards/margins": 0.2757399380207062, + "rewards/rejected": -0.2741301357746124, + "step": 272 + }, + { + "epoch": 0.5993413830954994, + "grad_norm": 3.419529438018799, + "learning_rate": 4.458028490219361e-06, + "logits/chosen": -2.3934783935546875, + "logits/rejected": -2.30633807182312, + "logps/chosen": -16.26495933532715, + "logps/rejected": -10.53646469116211, + "loss": 0.8953, + "rewards/accuracies": 0.6875, + "rewards/chosen": 0.11163412034511566, + "rewards/margins": 0.28469380736351013, + "rewards/rejected": -0.17305967211723328, + "step": 273 + }, + { + "epoch": 0.601536772777168, + "grad_norm": 3.859889268875122, + "learning_rate": 4.451985066691649e-06, + "logits/chosen": -2.3641891479492188, + "logits/rejected": -2.296032428741455, + "logps/chosen": -12.692946434020996, + "logps/rejected": -10.588760375976562, + "loss": 0.9138, + "rewards/accuracies": 0.625, + "rewards/chosen": 0.07239022850990295, + "rewards/margins": 0.19928386807441711, + "rewards/rejected": -0.12689363956451416, + "step": 274 + }, + { + "epoch": 0.6037321624588364, + "grad_norm": 3.1660425662994385, + "learning_rate": 4.445912279810401e-06, + "logits/chosen": -2.313688278198242, + "logits/rejected": -2.2848196029663086, + "logps/chosen": -18.03732681274414, + "logps/rejected": -13.442127227783203, + "loss": 0.9861, + "rewards/accuracies": 0.625, + "rewards/chosen": 0.0005273278802633286, + "rewards/margins": 0.19775639474391937, + "rewards/rejected": -0.1972290724515915, + "step": 275 + }, + { + "epoch": 0.605927552140505, + "grad_norm": 3.7765965461730957, + "learning_rate": 4.439810220927436e-06, + "logits/chosen": -2.4024770259857178, + "logits/rejected": -2.3250467777252197, + "logps/chosen": -16.64607810974121, + "logps/rejected": -14.17449951171875, + "loss": 0.9478, + "rewards/accuracies": 0.6875, + "rewards/chosen": -0.026453014463186264, + "rewards/margins": 0.30205315351486206, + "rewards/rejected": -0.3285061717033386, + "step": 276 + }, + { + "epoch": 0.6081229418221734, + "grad_norm": 4.055322170257568, + "learning_rate": 4.4336789818349105e-06, + "logits/chosen": -2.289247751235962, + "logits/rejected": -2.4132933616638184, + "logps/chosen": -19.57721519470215, + "logps/rejected": -11.495363235473633, + "loss": 0.8171, + "rewards/accuracies": 0.78125, + "rewards/chosen": 0.03967863321304321, + "rewards/margins": 0.4562482535839081, + "rewards/rejected": -0.41656967997550964, + "step": 277 + }, + { + "epoch": 0.610318331503842, + "grad_norm": 3.6051321029663086, + "learning_rate": 4.427518654763927e-06, + "logits/chosen": -2.358933448791504, + "logits/rejected": -2.4015445709228516, + "logps/chosen": -15.9168119430542, + "logps/rejected": -10.469045639038086, + "loss": 0.8713, + "rewards/accuracies": 0.78125, + "rewards/chosen": 0.08597777783870697, + "rewards/margins": 0.32388389110565186, + "rewards/rejected": -0.23790611326694489, + "step": 278 + }, + { + "epoch": 0.6125137211855104, + "grad_norm": 2.8481502532958984, + "learning_rate": 4.421329332383158e-06, + "logits/chosen": -2.40431547164917, + "logits/rejected": -2.313837766647339, + "logps/chosen": -16.356674194335938, + "logps/rejected": -11.987436294555664, + "loss": 0.9411, + "rewards/accuracies": 0.71875, + "rewards/chosen": 0.0003596879541873932, + "rewards/margins": 0.22185388207435608, + "rewards/rejected": -0.2214941829442978, + "step": 279 + }, + { + "epoch": 0.6147091108671789, + "grad_norm": 2.792403221130371, + "learning_rate": 4.415111107797445e-06, + "logits/chosen": -2.369375228881836, + "logits/rejected": -2.410369634628296, + "logps/chosen": -15.887426376342773, + "logps/rejected": -10.735150337219238, + "loss": 0.8151, + "rewards/accuracies": 0.71875, + "rewards/chosen": 0.1267678141593933, + "rewards/margins": 0.3933025598526001, + "rewards/rejected": -0.2665347456932068, + "step": 280 + }, + { + "epoch": 0.6169045005488474, + "grad_norm": 3.9273605346679688, + "learning_rate": 4.408864074546402e-06, + "logits/chosen": -2.300583839416504, + "logits/rejected": -2.2920844554901123, + "logps/chosen": -17.110836029052734, + "logps/rejected": -11.550253868103027, + "loss": 0.864, + "rewards/accuracies": 0.6875, + "rewards/chosen": -0.015616696327924728, + "rewards/margins": 0.36540573835372925, + "rewards/rejected": -0.3810224235057831, + "step": 281 + }, + { + "epoch": 0.6190998902305159, + "grad_norm": 3.407844066619873, + "learning_rate": 4.402588326603002e-06, + "logits/chosen": -2.3792104721069336, + "logits/rejected": -2.3388185501098633, + "logps/chosen": -15.377979278564453, + "logps/rejected": -13.580524444580078, + "loss": 1.0169, + "rewards/accuracies": 0.5625, + "rewards/chosen": -0.0523822084069252, + "rewards/margins": 0.22621504962444305, + "rewards/rejected": -0.27859723567962646, + "step": 282 + }, + { + "epoch": 0.6212952799121844, + "grad_norm": 2.764402151107788, + "learning_rate": 4.396283958372173e-06, + "logits/chosen": -2.3100624084472656, + "logits/rejected": -2.379979133605957, + "logps/chosen": -14.444868087768555, + "logps/rejected": -11.091350555419922, + "loss": 0.9187, + "rewards/accuracies": 0.5625, + "rewards/chosen": 0.0230597835034132, + "rewards/margins": 0.31620121002197266, + "rewards/rejected": -0.2931414246559143, + "step": 283 + }, + { + "epoch": 0.6234906695938529, + "grad_norm": 4.228127956390381, + "learning_rate": 4.38995106468937e-06, + "logits/chosen": -2.333906412124634, + "logits/rejected": -2.3636553287506104, + "logps/chosen": -14.673521041870117, + "logps/rejected": -9.182047843933105, + "loss": 0.821, + "rewards/accuracies": 0.75, + "rewards/chosen": 0.10073263198137283, + "rewards/margins": 0.3078424334526062, + "rewards/rejected": -0.20710980892181396, + "step": 284 + }, + { + "epoch": 0.6256860592755215, + "grad_norm": 3.735618829727173, + "learning_rate": 4.3835897408191515e-06, + "logits/chosen": -2.3141889572143555, + "logits/rejected": -2.357758045196533, + "logps/chosen": -16.62046241760254, + "logps/rejected": -10.231853485107422, + "loss": 0.841, + "rewards/accuracies": 0.71875, + "rewards/chosen": 0.0775555670261383, + "rewards/margins": 0.38503366708755493, + "rewards/rejected": -0.30747807025909424, + "step": 285 + }, + { + "epoch": 0.6278814489571899, + "grad_norm": 3.217437982559204, + "learning_rate": 4.377200082453748e-06, + "logits/chosen": -2.2870125770568848, + "logits/rejected": -2.3641347885131836, + "logps/chosen": -18.92882537841797, + "logps/rejected": -12.308082580566406, + "loss": 0.9247, + "rewards/accuracies": 0.75, + "rewards/chosen": 0.09097970277070999, + "rewards/margins": 0.43049687147140503, + "rewards/rejected": -0.33951717615127563, + "step": 286 + }, + { + "epoch": 0.6300768386388584, + "grad_norm": 2.8335301876068115, + "learning_rate": 4.370782185711618e-06, + "logits/chosen": -2.322899341583252, + "logits/rejected": -2.3178961277008057, + "logps/chosen": -13.624707221984863, + "logps/rejected": -12.783918380737305, + "loss": 0.9819, + "rewards/accuracies": 0.59375, + "rewards/chosen": 0.046398602426052094, + "rewards/margins": 0.20317067205905914, + "rewards/rejected": -0.15677204728126526, + "step": 287 + }, + { + "epoch": 0.6322722283205269, + "grad_norm": 4.198422431945801, + "learning_rate": 4.3643361471360045e-06, + "logits/chosen": -2.253892421722412, + "logits/rejected": -2.351269245147705, + "logps/chosen": -15.725739479064941, + "logps/rejected": -13.157400131225586, + "loss": 0.9817, + "rewards/accuracies": 0.6875, + "rewards/chosen": 0.0788668841123581, + "rewards/margins": 0.3620368242263794, + "rewards/rejected": -0.2831699252128601, + "step": 288 + }, + { + "epoch": 0.6344676180021954, + "grad_norm": 3.027604341506958, + "learning_rate": 4.357862063693486e-06, + "logits/chosen": -2.249347448348999, + "logits/rejected": -2.27540922164917, + "logps/chosen": -20.544876098632812, + "logps/rejected": -17.780059814453125, + "loss": 1.0549, + "rewards/accuracies": 0.84375, + "rewards/chosen": 0.07425278425216675, + "rewards/margins": 0.4821486473083496, + "rewards/rejected": -0.40789586305618286, + "step": 289 + }, + { + "epoch": 0.6366630076838639, + "grad_norm": 3.498201847076416, + "learning_rate": 4.351360032772512e-06, + "logits/chosen": -2.291285514831543, + "logits/rejected": -2.4294636249542236, + "logps/chosen": -16.40362548828125, + "logps/rejected": -9.61184310913086, + "loss": 0.8461, + "rewards/accuracies": 0.625, + "rewards/chosen": 0.058083947747945786, + "rewards/margins": 0.33970099687576294, + "rewards/rejected": -0.28161704540252686, + "step": 290 + }, + { + "epoch": 0.6388583973655324, + "grad_norm": 4.065576076507568, + "learning_rate": 4.344830152181941e-06, + "logits/chosen": -2.2679929733276367, + "logits/rejected": -2.359039306640625, + "logps/chosen": -17.067501068115234, + "logps/rejected": -11.277329444885254, + "loss": 0.9347, + "rewards/accuracies": 0.625, + "rewards/chosen": 0.01216885820031166, + "rewards/margins": 0.23538747429847717, + "rewards/rejected": -0.2232186198234558, + "step": 291 + }, + { + "epoch": 0.6410537870472008, + "grad_norm": 4.737266540527344, + "learning_rate": 4.338272520149572e-06, + "logits/chosen": -2.2955241203308105, + "logits/rejected": -2.382547616958618, + "logps/chosen": -17.920217514038086, + "logps/rejected": -15.063626289367676, + "loss": 1.0549, + "rewards/accuracies": 0.59375, + "rewards/chosen": -0.04992213100194931, + "rewards/margins": 0.10247353464365005, + "rewards/rejected": -0.15239566564559937, + "step": 292 + }, + { + "epoch": 0.6432491767288694, + "grad_norm": 4.056772708892822, + "learning_rate": 4.3316872353206595e-06, + "logits/chosen": -2.4128801822662354, + "logits/rejected": -2.3561244010925293, + "logps/chosen": -13.71163558959961, + "logps/rejected": -10.349092483520508, + "loss": 0.8511, + "rewards/accuracies": 0.75, + "rewards/chosen": 0.007009743247181177, + "rewards/margins": 0.2926795482635498, + "rewards/rejected": -0.28566980361938477, + "step": 293 + }, + { + "epoch": 0.6454445664105378, + "grad_norm": 3.3059873580932617, + "learning_rate": 4.325074396756437e-06, + "logits/chosen": -2.3915586471557617, + "logits/rejected": -2.242203712463379, + "logps/chosen": -17.347070693969727, + "logps/rejected": -10.363704681396484, + "loss": 0.8055, + "rewards/accuracies": 0.84375, + "rewards/chosen": 0.05488693714141846, + "rewards/margins": 0.3893585205078125, + "rewards/rejected": -0.33447155356407166, + "step": 294 + }, + { + "epoch": 0.6476399560922064, + "grad_norm": 3.126178026199341, + "learning_rate": 4.318434103932622e-06, + "logits/chosen": -2.377377510070801, + "logits/rejected": -2.3431365489959717, + "logps/chosen": -17.024227142333984, + "logps/rejected": -11.137166023254395, + "loss": 0.8691, + "rewards/accuracies": 0.71875, + "rewards/chosen": 0.11715853214263916, + "rewards/margins": 0.39744699001312256, + "rewards/rejected": -0.280288428068161, + "step": 295 + }, + { + "epoch": 0.6498353457738749, + "grad_norm": 3.4686014652252197, + "learning_rate": 4.3117664567379235e-06, + "logits/chosen": -2.3375179767608643, + "logits/rejected": -2.2733144760131836, + "logps/chosen": -17.800548553466797, + "logps/rejected": -10.393657684326172, + "loss": 0.8631, + "rewards/accuracies": 0.65625, + "rewards/chosen": 0.01607682555913925, + "rewards/margins": 0.31594520807266235, + "rewards/rejected": -0.2998683452606201, + "step": 296 + }, + { + "epoch": 0.6520307354555434, + "grad_norm": 3.257317304611206, + "learning_rate": 4.305071555472534e-06, + "logits/chosen": -2.3289194107055664, + "logits/rejected": -2.4071145057678223, + "logps/chosen": -14.666779518127441, + "logps/rejected": -9.785927772521973, + "loss": 0.8539, + "rewards/accuracies": 0.71875, + "rewards/chosen": 0.029782062396407127, + "rewards/margins": 0.368835985660553, + "rewards/rejected": -0.3390539288520813, + "step": 297 + }, + { + "epoch": 0.6542261251372119, + "grad_norm": 3.4349687099456787, + "learning_rate": 4.2983495008466285e-06, + "logits/chosen": -2.3456382751464844, + "logits/rejected": -2.421517848968506, + "logps/chosen": -16.462114334106445, + "logps/rejected": -14.361862182617188, + "loss": 0.9191, + "rewards/accuracies": 0.75, + "rewards/chosen": 0.03672860935330391, + "rewards/margins": 0.4619566798210144, + "rewards/rejected": -0.425228089094162, + "step": 298 + }, + { + "epoch": 0.6564215148188803, + "grad_norm": 4.040913105010986, + "learning_rate": 4.29160039397884e-06, + "logits/chosen": -2.3418679237365723, + "logits/rejected": -2.2938270568847656, + "logps/chosen": -11.77094554901123, + "logps/rejected": -13.91958999633789, + "loss": 0.9844, + "rewards/accuracies": 0.65625, + "rewards/chosen": 0.06349484622478485, + "rewards/margins": 0.26532605290412903, + "rewards/rejected": -0.20183119177818298, + "step": 299 + }, + { + "epoch": 0.6586169045005489, + "grad_norm": 4.860092639923096, + "learning_rate": 4.284824336394748e-06, + "logits/chosen": -2.3495969772338867, + "logits/rejected": -2.376173496246338, + "logps/chosen": -14.834826469421387, + "logps/rejected": -13.061057090759277, + "loss": 0.9285, + "rewards/accuracies": 0.78125, + "rewards/chosen": 0.03790712356567383, + "rewards/margins": 0.3323096036911011, + "rewards/rejected": -0.29440245032310486, + "step": 300 + }, + { + "epoch": 0.6608122941822173, + "grad_norm": 3.8926429748535156, + "learning_rate": 4.278021430025343e-06, + "logits/chosen": -2.4280104637145996, + "logits/rejected": -2.384596586227417, + "logps/chosen": -15.310932159423828, + "logps/rejected": -9.883270263671875, + "loss": 0.8233, + "rewards/accuracies": 0.78125, + "rewards/chosen": 0.03349144011735916, + "rewards/margins": 0.47141653299331665, + "rewards/rejected": -0.4379251003265381, + "step": 301 + }, + { + "epoch": 0.6630076838638859, + "grad_norm": 4.136510848999023, + "learning_rate": 4.2711917772055e-06, + "logits/chosen": -2.349405288696289, + "logits/rejected": -2.4378554821014404, + "logps/chosen": -20.67293930053711, + "logps/rejected": -17.613826751708984, + "loss": 1.0321, + "rewards/accuracies": 0.6875, + "rewards/chosen": 0.02313861809670925, + "rewards/margins": 0.3090207576751709, + "rewards/rejected": -0.2858821451663971, + "step": 302 + }, + { + "epoch": 0.6652030735455543, + "grad_norm": 3.6378984451293945, + "learning_rate": 4.264335480672433e-06, + "logits/chosen": -2.339221954345703, + "logits/rejected": -2.355464458465576, + "logps/chosen": -14.879115104675293, + "logps/rejected": -11.936575889587402, + "loss": 0.9292, + "rewards/accuracies": 0.75, + "rewards/chosen": 0.11328764259815216, + "rewards/margins": 0.28763386607170105, + "rewards/rejected": -0.1743462085723877, + "step": 303 + }, + { + "epoch": 0.6673984632272228, + "grad_norm": 4.390617847442627, + "learning_rate": 4.257452643564155e-06, + "logits/chosen": -2.3608078956604004, + "logits/rejected": -2.2916340827941895, + "logps/chosen": -16.132726669311523, + "logps/rejected": -11.27886962890625, + "loss": 0.8445, + "rewards/accuracies": 0.78125, + "rewards/chosen": 0.013227814808487892, + "rewards/margins": 0.38369402289390564, + "rewards/rejected": -0.3704661726951599, + "step": 304 + }, + { + "epoch": 0.6695938529088913, + "grad_norm": 4.490567684173584, + "learning_rate": 4.250543369417921e-06, + "logits/chosen": -2.334174156188965, + "logits/rejected": -2.340644359588623, + "logps/chosen": -13.944917678833008, + "logps/rejected": -10.003447532653809, + "loss": 0.9194, + "rewards/accuracies": 0.625, + "rewards/chosen": 0.014632996171712875, + "rewards/margins": 0.17052799463272095, + "rewards/rejected": -0.15589500963687897, + "step": 305 + }, + { + "epoch": 0.6717892425905598, + "grad_norm": 3.990586042404175, + "learning_rate": 4.2436077621686785e-06, + "logits/chosen": -2.324025869369507, + "logits/rejected": -2.400771379470825, + "logps/chosen": -18.718528747558594, + "logps/rejected": -9.990530014038086, + "loss": 0.7944, + "rewards/accuracies": 0.78125, + "rewards/chosen": 0.056847233325242996, + "rewards/margins": 0.49278268218040466, + "rewards/rejected": -0.43593543767929077, + "step": 306 + }, + { + "epoch": 0.6739846322722283, + "grad_norm": 4.3013505935668945, + "learning_rate": 4.236645926147493e-06, + "logits/chosen": -2.38020920753479, + "logits/rejected": -2.3586490154266357, + "logps/chosen": -13.938803672790527, + "logps/rejected": -12.504024505615234, + "loss": 0.9618, + "rewards/accuracies": 0.5625, + "rewards/chosen": 0.051238127052783966, + "rewards/margins": 0.1698467880487442, + "rewards/rejected": -0.11860863119363785, + "step": 307 + }, + { + "epoch": 0.6761800219538968, + "grad_norm": 4.09398078918457, + "learning_rate": 4.22965796607999e-06, + "logits/chosen": -2.253674268722534, + "logits/rejected": -2.3258652687072754, + "logps/chosen": -20.03459930419922, + "logps/rejected": -14.505156517028809, + "loss": 0.9795, + "rewards/accuracies": 0.625, + "rewards/chosen": 0.04597073420882225, + "rewards/margins": 0.39578795433044434, + "rewards/rejected": -0.3498172163963318, + "step": 308 + }, + { + "epoch": 0.6783754116355654, + "grad_norm": 3.907806634902954, + "learning_rate": 4.22264398708477e-06, + "logits/chosen": -2.3766307830810547, + "logits/rejected": -2.3582723140716553, + "logps/chosen": -14.181205749511719, + "logps/rejected": -10.14924430847168, + "loss": 0.9015, + "rewards/accuracies": 0.59375, + "rewards/chosen": 0.028858356177806854, + "rewards/margins": 0.2582564055919647, + "rewards/rejected": -0.22939805686473846, + "step": 309 + }, + { + "epoch": 0.6805708013172338, + "grad_norm": 3.5894784927368164, + "learning_rate": 4.215604094671835e-06, + "logits/chosen": -2.3014726638793945, + "logits/rejected": -2.3518590927124023, + "logps/chosen": -14.997878074645996, + "logps/rejected": -10.144777297973633, + "loss": 0.7828, + "rewards/accuracies": 0.84375, + "rewards/chosen": 0.04976440966129303, + "rewards/margins": 0.513340950012207, + "rewards/rejected": -0.4635765254497528, + "step": 310 + }, + { + "epoch": 0.6827661909989023, + "grad_norm": 4.483096122741699, + "learning_rate": 4.208538394740993e-06, + "logits/chosen": -2.2711143493652344, + "logits/rejected": -2.3518011569976807, + "logps/chosen": -17.233510971069336, + "logps/rejected": -12.07126235961914, + "loss": 0.9666, + "rewards/accuracies": 0.71875, + "rewards/chosen": 0.055232856422662735, + "rewards/margins": 0.24555866420269012, + "rewards/rejected": -0.1903257966041565, + "step": 311 + }, + { + "epoch": 0.6849615806805708, + "grad_norm": 4.8697967529296875, + "learning_rate": 4.201446993580277e-06, + "logits/chosen": -2.255201816558838, + "logits/rejected": -2.418351650238037, + "logps/chosen": -16.337543487548828, + "logps/rejected": -10.240762710571289, + "loss": 0.9028, + "rewards/accuracies": 0.6875, + "rewards/chosen": -0.04591435566544533, + "rewards/margins": 0.3530338406562805, + "rewards/rejected": -0.39894816279411316, + "step": 312 + }, + { + "epoch": 0.6871569703622393, + "grad_norm": 4.00418758392334, + "learning_rate": 4.194329997864331e-06, + "logits/chosen": -2.4171342849731445, + "logits/rejected": -2.378293037414551, + "logps/chosen": -18.34632110595703, + "logps/rejected": -12.369073867797852, + "loss": 0.8811, + "rewards/accuracies": 0.71875, + "rewards/chosen": 0.001896400935947895, + "rewards/margins": 0.48826920986175537, + "rewards/rejected": -0.48637276887893677, + "step": 313 + }, + { + "epoch": 0.6893523600439078, + "grad_norm": 4.626528739929199, + "learning_rate": 4.18718751465282e-06, + "logits/chosen": -2.329511880874634, + "logits/rejected": -2.3417491912841797, + "logps/chosen": -14.74193000793457, + "logps/rejected": -10.454490661621094, + "loss": 0.8867, + "rewards/accuracies": 0.65625, + "rewards/chosen": 0.04431447759270668, + "rewards/margins": 0.42057913541793823, + "rewards/rejected": -0.37626469135284424, + "step": 314 + }, + { + "epoch": 0.6915477497255763, + "grad_norm": 3.7983834743499756, + "learning_rate": 4.180019651388807e-06, + "logits/chosen": -2.2895097732543945, + "logits/rejected": -2.3450984954833984, + "logps/chosen": -15.74697208404541, + "logps/rejected": -11.167444229125977, + "loss": 0.8818, + "rewards/accuracies": 0.75, + "rewards/chosen": 0.03174567595124245, + "rewards/margins": 0.35864511132240295, + "rewards/rejected": -0.3268994688987732, + "step": 315 + }, + { + "epoch": 0.6937431394072447, + "grad_norm": 3.4797542095184326, + "learning_rate": 4.172826515897146e-06, + "logits/chosen": -2.311906337738037, + "logits/rejected": -2.3090274333953857, + "logps/chosen": -13.851794242858887, + "logps/rejected": -12.625475883483887, + "loss": 0.9461, + "rewards/accuracies": 0.71875, + "rewards/chosen": 0.017236392945051193, + "rewards/margins": 0.2493745982646942, + "rewards/rejected": -0.23213821649551392, + "step": 316 + }, + { + "epoch": 0.6959385290889133, + "grad_norm": 4.422733306884766, + "learning_rate": 4.165608216382857e-06, + "logits/chosen": -2.4253227710723877, + "logits/rejected": -2.3780248165130615, + "logps/chosen": -16.252634048461914, + "logps/rejected": -13.770538330078125, + "loss": 0.9458, + "rewards/accuracies": 0.59375, + "rewards/chosen": 0.06439048796892166, + "rewards/margins": 0.30147218704223633, + "rewards/rejected": -0.23708167672157288, + "step": 317 + }, + { + "epoch": 0.6981339187705817, + "grad_norm": 3.7859911918640137, + "learning_rate": 4.158364861429493e-06, + "logits/chosen": -2.322923421859741, + "logits/rejected": -2.2539587020874023, + "logps/chosen": -18.831212997436523, + "logps/rejected": -11.069580078125, + "loss": 0.8346, + "rewards/accuracies": 0.71875, + "rewards/chosen": 0.03058037906885147, + "rewards/margins": 0.34845486283302307, + "rewards/rejected": -0.3178745210170746, + "step": 318 + }, + { + "epoch": 0.7003293084522503, + "grad_norm": 3.731484889984131, + "learning_rate": 4.151096559997519e-06, + "logits/chosen": -2.3627357482910156, + "logits/rejected": -2.3674192428588867, + "logps/chosen": -15.868612289428711, + "logps/rejected": -10.797269821166992, + "loss": 0.8928, + "rewards/accuracies": 0.71875, + "rewards/chosen": 0.02595650777220726, + "rewards/margins": 0.3343779742717743, + "rewards/rejected": -0.30842143297195435, + "step": 319 + }, + { + "epoch": 0.7025246981339188, + "grad_norm": 3.9332573413848877, + "learning_rate": 4.14380342142266e-06, + "logits/chosen": -2.3744566440582275, + "logits/rejected": -2.309058904647827, + "logps/chosen": -12.979777336120605, + "logps/rejected": -12.636903762817383, + "loss": 0.9292, + "rewards/accuracies": 0.75, + "rewards/chosen": 0.08498867601156235, + "rewards/margins": 0.41657087206840515, + "rewards/rejected": -0.3315821886062622, + "step": 320 + }, + { + "epoch": 0.7047200878155873, + "grad_norm": 5.405649185180664, + "learning_rate": 4.1364855554142604e-06, + "logits/chosen": -2.3514606952667236, + "logits/rejected": -2.2680933475494385, + "logps/chosen": -15.088423728942871, + "logps/rejected": -11.752235412597656, + "loss": 0.881, + "rewards/accuracies": 0.6875, + "rewards/chosen": -0.05181232839822769, + "rewards/margins": 0.3780396580696106, + "rewards/rejected": -0.4298520088195801, + "step": 321 + }, + { + "epoch": 0.7069154774972558, + "grad_norm": 4.955357551574707, + "learning_rate": 4.129143072053639e-06, + "logits/chosen": -2.3219053745269775, + "logits/rejected": -2.38175630569458, + "logps/chosen": -15.868922233581543, + "logps/rejected": -10.88913345336914, + "loss": 0.8665, + "rewards/accuracies": 0.6875, + "rewards/chosen": -0.052466217428445816, + "rewards/margins": 0.5013300776481628, + "rewards/rejected": -0.5537962913513184, + "step": 322 + }, + { + "epoch": 0.7091108671789242, + "grad_norm": 4.574494361877441, + "learning_rate": 4.121776081792426e-06, + "logits/chosen": -2.3393287658691406, + "logits/rejected": -2.381781578063965, + "logps/chosen": -18.754756927490234, + "logps/rejected": -14.181581497192383, + "loss": 0.8978, + "rewards/accuracies": 0.75, + "rewards/chosen": 0.04621858894824982, + "rewards/margins": 0.47777223587036133, + "rewards/rejected": -0.4315536618232727, + "step": 323 + }, + { + "epoch": 0.7113062568605928, + "grad_norm": 4.758660316467285, + "learning_rate": 4.114384695450906e-06, + "logits/chosen": -2.3381195068359375, + "logits/rejected": -2.3528432846069336, + "logps/chosen": -16.682260513305664, + "logps/rejected": -10.463571548461914, + "loss": 0.8577, + "rewards/accuracies": 0.78125, + "rewards/chosen": 0.05263161659240723, + "rewards/margins": 0.49307867884635925, + "rewards/rejected": -0.440447062253952, + "step": 324 + }, + { + "epoch": 0.7135016465422612, + "grad_norm": 5.131170749664307, + "learning_rate": 4.106969024216348e-06, + "logits/chosen": -2.304837703704834, + "logits/rejected": -2.3518829345703125, + "logps/chosen": -17.55645179748535, + "logps/rejected": -11.162080764770508, + "loss": 0.7799, + "rewards/accuracies": 0.90625, + "rewards/chosen": 0.09621885418891907, + "rewards/margins": 0.5682495832443237, + "rewards/rejected": -0.4720306396484375, + "step": 325 + }, + { + "epoch": 0.7156970362239298, + "grad_norm": 4.7023844718933105, + "learning_rate": 4.099529179641337e-06, + "logits/chosen": -2.31040358543396, + "logits/rejected": -2.3163864612579346, + "logps/chosen": -14.483461380004883, + "logps/rejected": -11.768637657165527, + "loss": 0.9377, + "rewards/accuracies": 0.5625, + "rewards/chosen": 0.0021479413844645023, + "rewards/margins": 0.3606518507003784, + "rewards/rejected": -0.35850390791893005, + "step": 326 + }, + { + "epoch": 0.7178924259055982, + "grad_norm": 3.8112165927886963, + "learning_rate": 4.09206527364209e-06, + "logits/chosen": -2.382301092147827, + "logits/rejected": -2.365294933319092, + "logps/chosen": -16.85517120361328, + "logps/rejected": -12.415393829345703, + "loss": 0.9548, + "rewards/accuracies": 0.59375, + "rewards/chosen": -0.035888537764549255, + "rewards/margins": 0.31409651041030884, + "rewards/rejected": -0.3499850034713745, + "step": 327 + }, + { + "epoch": 0.7200878155872668, + "grad_norm": 3.9805612564086914, + "learning_rate": 4.084577418496775e-06, + "logits/chosen": -2.3480184078216553, + "logits/rejected": -2.3130125999450684, + "logps/chosen": -13.996624946594238, + "logps/rejected": -10.171143531799316, + "loss": 0.7951, + "rewards/accuracies": 0.84375, + "rewards/chosen": 0.09903817623853683, + "rewards/margins": 0.4770241379737854, + "rewards/rejected": -0.37798595428466797, + "step": 328 + }, + { + "epoch": 0.7222832052689352, + "grad_norm": 4.388843536376953, + "learning_rate": 4.0770657268438285e-06, + "logits/chosen": -2.250278949737549, + "logits/rejected": -2.3578574657440186, + "logps/chosen": -17.343597412109375, + "logps/rejected": -7.6419196128845215, + "loss": 0.721, + "rewards/accuracies": 0.84375, + "rewards/chosen": 0.09454993903636932, + "rewards/margins": 0.7629677057266235, + "rewards/rejected": -0.6684178113937378, + "step": 329 + }, + { + "epoch": 0.7244785949506037, + "grad_norm": 6.914097309112549, + "learning_rate": 4.069530311680247e-06, + "logits/chosen": -2.3882980346679688, + "logits/rejected": -2.374696969985962, + "logps/chosen": -15.285945892333984, + "logps/rejected": -11.684650421142578, + "loss": 0.9118, + "rewards/accuracies": 0.71875, + "rewards/chosen": -0.05909008905291557, + "rewards/margins": 0.3279057443141937, + "rewards/rejected": -0.3869958221912384, + "step": 330 + }, + { + "epoch": 0.7266739846322722, + "grad_norm": 5.314255237579346, + "learning_rate": 4.0619712863599005e-06, + "logits/chosen": -2.278480291366577, + "logits/rejected": -2.2412896156311035, + "logps/chosen": -15.4207124710083, + "logps/rejected": -11.311100959777832, + "loss": 0.8939, + "rewards/accuracies": 0.625, + "rewards/chosen": 0.008756492286920547, + "rewards/margins": 0.43041741847991943, + "rewards/rejected": -0.4216609001159668, + "step": 331 + }, + { + "epoch": 0.7288693743139407, + "grad_norm": 4.677224636077881, + "learning_rate": 4.054388764591822e-06, + "logits/chosen": -2.344648838043213, + "logits/rejected": -2.395387887954712, + "logps/chosen": -17.03143882751465, + "logps/rejected": -10.78937816619873, + "loss": 0.8664, + "rewards/accuracies": 0.71875, + "rewards/chosen": -0.012906208634376526, + "rewards/margins": 0.3622843325138092, + "rewards/rejected": -0.3751905858516693, + "step": 332 + }, + { + "epoch": 0.7310647639956093, + "grad_norm": 4.577656269073486, + "learning_rate": 4.046782860438497e-06, + "logits/chosen": -2.3226308822631836, + "logits/rejected": -2.265408992767334, + "logps/chosen": -15.200016021728516, + "logps/rejected": -10.772403717041016, + "loss": 0.9198, + "rewards/accuracies": 0.71875, + "rewards/chosen": -0.07568801939487457, + "rewards/margins": 0.40591323375701904, + "rewards/rejected": -0.4816012382507324, + "step": 333 + }, + { + "epoch": 0.7332601536772777, + "grad_norm": 4.119585990905762, + "learning_rate": 4.039153688314146e-06, + "logits/chosen": -2.3442039489746094, + "logits/rejected": -2.359684467315674, + "logps/chosen": -18.90411376953125, + "logps/rejected": -10.934317588806152, + "loss": 0.7644, + "rewards/accuracies": 0.90625, + "rewards/chosen": 0.03230892866849899, + "rewards/margins": 0.6441935300827026, + "rewards/rejected": -0.611884593963623, + "step": 334 + }, + { + "epoch": 0.7354555433589463, + "grad_norm": 5.220119476318359, + "learning_rate": 4.031501362983007e-06, + "logits/chosen": -2.297410726547241, + "logits/rejected": -2.481637716293335, + "logps/chosen": -15.908681869506836, + "logps/rejected": -11.298460006713867, + "loss": 0.9041, + "rewards/accuracies": 0.625, + "rewards/chosen": 0.017940856516361237, + "rewards/margins": 0.29080140590667725, + "rewards/rejected": -0.272860586643219, + "step": 335 + }, + { + "epoch": 0.7376509330406147, + "grad_norm": 4.522210597991943, + "learning_rate": 4.023825999557608e-06, + "logits/chosen": -2.3637242317199707, + "logits/rejected": -2.355468273162842, + "logps/chosen": -17.870847702026367, + "logps/rejected": -12.528518676757812, + "loss": 0.9398, + "rewards/accuracies": 0.65625, + "rewards/chosen": -0.06956735253334045, + "rewards/margins": 0.298219233751297, + "rewards/rejected": -0.36778658628463745, + "step": 336 + }, + { + "epoch": 0.7398463227222832, + "grad_norm": 6.380414009094238, + "learning_rate": 4.016127713497034e-06, + "logits/chosen": -2.298593521118164, + "logits/rejected": -2.343686103820801, + "logps/chosen": -16.275299072265625, + "logps/rejected": -12.0189847946167, + "loss": 0.9632, + "rewards/accuracies": 0.75, + "rewards/chosen": -0.0297526977956295, + "rewards/margins": 0.38488292694091797, + "rewards/rejected": -0.4146355986595154, + "step": 337 + }, + { + "epoch": 0.7420417124039517, + "grad_norm": 5.236372470855713, + "learning_rate": 4.00840662060519e-06, + "logits/chosen": -2.359950542449951, + "logits/rejected": -2.2415010929107666, + "logps/chosen": -24.300395965576172, + "logps/rejected": -15.511738777160645, + "loss": 1.0485, + "rewards/accuracies": 0.53125, + "rewards/chosen": -0.06561987102031708, + "rewards/margins": 0.27457380294799805, + "rewards/rejected": -0.3401937186717987, + "step": 338 + }, + { + "epoch": 0.7442371020856202, + "grad_norm": 4.642820358276367, + "learning_rate": 4.000662837029062e-06, + "logits/chosen": -2.4099583625793457, + "logits/rejected": -2.354393243789673, + "logps/chosen": -22.309005737304688, + "logps/rejected": -11.256224632263184, + "loss": 0.7581, + "rewards/accuracies": 0.78125, + "rewards/chosen": 0.010493040084838867, + "rewards/margins": 0.7678316831588745, + "rewards/rejected": -0.7573386430740356, + "step": 339 + }, + { + "epoch": 0.7464324917672887, + "grad_norm": 3.9868204593658447, + "learning_rate": 3.992896479256966e-06, + "logits/chosen": -2.310539722442627, + "logits/rejected": -2.306994915008545, + "logps/chosen": -18.193031311035156, + "logps/rejected": -9.452789306640625, + "loss": 0.7584, + "rewards/accuracies": 0.75, + "rewards/chosen": 0.021110711619257927, + "rewards/margins": 0.8151036500930786, + "rewards/rejected": -0.7939929366111755, + "step": 340 + }, + { + "epoch": 0.7486278814489572, + "grad_norm": 4.947543621063232, + "learning_rate": 3.985107664116798e-06, + "logits/chosen": -2.3655166625976562, + "logits/rejected": -2.3199362754821777, + "logps/chosen": -16.473398208618164, + "logps/rejected": -9.421045303344727, + "loss": 0.8089, + "rewards/accuracies": 0.71875, + "rewards/chosen": 0.05412997305393219, + "rewards/margins": 0.39927852153778076, + "rewards/rejected": -0.345148503780365, + "step": 341 + }, + { + "epoch": 0.7508232711306256, + "grad_norm": 4.09507942199707, + "learning_rate": 3.977296508774278e-06, + "logits/chosen": -2.2579798698425293, + "logits/rejected": -2.3313004970550537, + "logps/chosen": -17.961259841918945, + "logps/rejected": -8.732832908630371, + "loss": 0.7648, + "rewards/accuracies": 0.71875, + "rewards/chosen": 0.051630470901727676, + "rewards/margins": 0.7588289976119995, + "rewards/rejected": -0.7071985006332397, + "step": 342 + }, + { + "epoch": 0.7530186608122942, + "grad_norm": 6.222644329071045, + "learning_rate": 3.969463130731183e-06, + "logits/chosen": -2.3719406127929688, + "logits/rejected": -2.3723111152648926, + "logps/chosen": -18.259965896606445, + "logps/rejected": -10.765856742858887, + "loss": 0.867, + "rewards/accuracies": 0.75, + "rewards/chosen": 0.09137167036533356, + "rewards/margins": 0.42636412382125854, + "rewards/rejected": -0.3349924087524414, + "step": 343 + }, + { + "epoch": 0.7552140504939627, + "grad_norm": 5.629691123962402, + "learning_rate": 3.9616076478235835e-06, + "logits/chosen": -2.3341000080108643, + "logits/rejected": -2.320246696472168, + "logps/chosen": -14.3938570022583, + "logps/rejected": -9.880708694458008, + "loss": 0.8374, + "rewards/accuracies": 0.8125, + "rewards/chosen": 0.06494603306055069, + "rewards/margins": 0.4671281576156616, + "rewards/rejected": -0.4021821916103363, + "step": 344 + }, + { + "epoch": 0.7574094401756312, + "grad_norm": 3.4992260932922363, + "learning_rate": 3.953730178220067e-06, + "logits/chosen": -2.349565267562866, + "logits/rejected": -2.394681930541992, + "logps/chosen": -16.43452262878418, + "logps/rejected": -11.140144348144531, + "loss": 0.8392, + "rewards/accuracies": 0.84375, + "rewards/chosen": 0.0837668776512146, + "rewards/margins": 0.46383053064346313, + "rewards/rejected": -0.38006365299224854, + "step": 345 + }, + { + "epoch": 0.7596048298572997, + "grad_norm": 5.090843677520752, + "learning_rate": 3.945830840419966e-06, + "logits/chosen": -2.3477964401245117, + "logits/rejected": -2.3646602630615234, + "logps/chosen": -16.911046981811523, + "logps/rejected": -8.938126564025879, + "loss": 0.8339, + "rewards/accuracies": 0.71875, + "rewards/chosen": -0.07053576409816742, + "rewards/margins": 0.6060460805892944, + "rewards/rejected": -0.676581859588623, + "step": 346 + }, + { + "epoch": 0.7618002195389681, + "grad_norm": 5.51276159286499, + "learning_rate": 3.937909753251572e-06, + "logits/chosen": -2.2657523155212402, + "logits/rejected": -2.330665111541748, + "logps/chosen": -16.019929885864258, + "logps/rejected": -9.611878395080566, + "loss": 0.8075, + "rewards/accuracies": 0.75, + "rewards/chosen": 0.07287417352199554, + "rewards/margins": 0.5842018127441406, + "rewards/rejected": -0.5113277435302734, + "step": 347 + }, + { + "epoch": 0.7639956092206367, + "grad_norm": 3.379443645477295, + "learning_rate": 3.929967035870346e-06, + "logits/chosen": -2.3243355751037598, + "logits/rejected": -2.2666611671447754, + "logps/chosen": -15.123114585876465, + "logps/rejected": -10.203433990478516, + "loss": 0.8297, + "rewards/accuracies": 0.75, + "rewards/chosen": 0.11292536556720734, + "rewards/margins": 0.45466747879981995, + "rewards/rejected": -0.3417420983314514, + "step": 348 + }, + { + "epoch": 0.7661909989023051, + "grad_norm": 8.115785598754883, + "learning_rate": 3.92200280775713e-06, + "logits/chosen": -2.35408878326416, + "logits/rejected": -2.3558552265167236, + "logps/chosen": -18.513586044311523, + "logps/rejected": -11.50524616241455, + "loss": 0.9296, + "rewards/accuracies": 0.6875, + "rewards/chosen": -0.06223009154200554, + "rewards/margins": 0.28361374139785767, + "rewards/rejected": -0.3458438515663147, + "step": 349 + }, + { + "epoch": 0.7683863885839737, + "grad_norm": 4.38709831237793, + "learning_rate": 3.914017188716347e-06, + "logits/chosen": -2.296597480773926, + "logits/rejected": -2.41675066947937, + "logps/chosen": -16.949928283691406, + "logps/rejected": -11.099538803100586, + "loss": 0.8672, + "rewards/accuracies": 0.65625, + "rewards/chosen": 0.11809515953063965, + "rewards/margins": 0.43611082434654236, + "rewards/rejected": -0.3180156648159027, + "step": 350 + }, + { + "epoch": 0.7705817782656421, + "grad_norm": 5.365295886993408, + "learning_rate": 3.9060102988742e-06, + "logits/chosen": -2.2918920516967773, + "logits/rejected": -2.329606056213379, + "logps/chosen": -18.76449203491211, + "logps/rejected": -12.69644546508789, + "loss": 0.8634, + "rewards/accuracies": 0.78125, + "rewards/chosen": -0.009303185157477856, + "rewards/margins": 0.6541052460670471, + "rewards/rejected": -0.6634083986282349, + "step": 351 + }, + { + "epoch": 0.7727771679473107, + "grad_norm": 3.892143964767456, + "learning_rate": 3.897982258676867e-06, + "logits/chosen": -2.3476452827453613, + "logits/rejected": -2.373446226119995, + "logps/chosen": -19.90625, + "logps/rejected": -12.041308403015137, + "loss": 0.8741, + "rewards/accuracies": 0.625, + "rewards/chosen": 0.06443409621715546, + "rewards/margins": 0.49310898780822754, + "rewards/rejected": -0.4286748766899109, + "step": 352 + }, + { + "epoch": 0.7749725576289791, + "grad_norm": 3.8393607139587402, + "learning_rate": 3.889933188888684e-06, + "logits/chosen": -2.273186206817627, + "logits/rejected": -2.268983840942383, + "logps/chosen": -14.942765235900879, + "logps/rejected": -10.225128173828125, + "loss": 0.8367, + "rewards/accuracies": 0.75, + "rewards/chosen": 0.13203339278697968, + "rewards/margins": 0.38842177391052246, + "rewards/rejected": -0.2563883662223816, + "step": 353 + }, + { + "epoch": 0.7771679473106476, + "grad_norm": 5.566308975219727, + "learning_rate": 3.881863210590332e-06, + "logits/chosen": -2.408592700958252, + "logits/rejected": -2.4445698261260986, + "logps/chosen": -16.196022033691406, + "logps/rejected": -10.760068893432617, + "loss": 0.8831, + "rewards/accuracies": 0.71875, + "rewards/chosen": -0.021768394857645035, + "rewards/margins": 0.4853077232837677, + "rewards/rejected": -0.5070760846138, + "step": 354 + }, + { + "epoch": 0.7793633369923162, + "grad_norm": 4.365970134735107, + "learning_rate": 3.8737724451770155e-06, + "logits/chosen": -2.3492016792297363, + "logits/rejected": -2.309647560119629, + "logps/chosen": -19.536739349365234, + "logps/rejected": -12.805992126464844, + "loss": 0.9172, + "rewards/accuracies": 0.78125, + "rewards/chosen": -0.07249265909194946, + "rewards/margins": 0.5484910011291504, + "rewards/rejected": -0.6209836602210999, + "step": 355 + }, + { + "epoch": 0.7815587266739846, + "grad_norm": 4.709788799285889, + "learning_rate": 3.865661014356635e-06, + "logits/chosen": -2.3180618286132812, + "logits/rejected": -2.2893731594085693, + "logps/chosen": -20.79785919189453, + "logps/rejected": -13.704549789428711, + "loss": 0.8771, + "rewards/accuracies": 0.8125, + "rewards/chosen": 0.12575146555900574, + "rewards/margins": 0.5705665946006775, + "rewards/rejected": -0.44481509923934937, + "step": 356 + }, + { + "epoch": 0.7837541163556532, + "grad_norm": 4.544501304626465, + "learning_rate": 3.857529040147959e-06, + "logits/chosen": -2.3178446292877197, + "logits/rejected": -2.2923343181610107, + "logps/chosen": -17.59714126586914, + "logps/rejected": -10.660222053527832, + "loss": 0.7999, + "rewards/accuracies": 0.8125, + "rewards/chosen": 0.03826873004436493, + "rewards/margins": 0.501665472984314, + "rewards/rejected": -0.4633967876434326, + "step": 357 + }, + { + "epoch": 0.7859495060373216, + "grad_norm": 5.793648719787598, + "learning_rate": 3.849376644878783e-06, + "logits/chosen": -2.2837600708007812, + "logits/rejected": -2.2606236934661865, + "logps/chosen": -19.073780059814453, + "logps/rejected": -13.95520305633545, + "loss": 0.9672, + "rewards/accuracies": 0.71875, + "rewards/chosen": 0.04491977393627167, + "rewards/margins": 0.4374791979789734, + "rewards/rejected": -0.3925594091415405, + "step": 358 + }, + { + "epoch": 0.7881448957189902, + "grad_norm": 7.747844219207764, + "learning_rate": 3.841203951184095e-06, + "logits/chosen": -2.341078281402588, + "logits/rejected": -2.320150852203369, + "logps/chosen": -17.046415328979492, + "logps/rejected": -14.675268173217773, + "loss": 1.0073, + "rewards/accuracies": 0.65625, + "rewards/chosen": -0.12396389245986938, + "rewards/margins": 0.3983825445175171, + "rewards/rejected": -0.5223464369773865, + "step": 359 + }, + { + "epoch": 0.7903402854006586, + "grad_norm": 6.052748680114746, + "learning_rate": 3.833011082004229e-06, + "logits/chosen": -2.2902965545654297, + "logits/rejected": -2.296034336090088, + "logps/chosen": -18.75448989868164, + "logps/rejected": -11.371520042419434, + "loss": 0.7632, + "rewards/accuracies": 0.9375, + "rewards/chosen": 0.03268669173121452, + "rewards/margins": 0.6433117389678955, + "rewards/rejected": -0.6106250286102295, + "step": 360 + }, + { + "epoch": 0.7925356750823271, + "grad_norm": 5.396251678466797, + "learning_rate": 3.824798160583012e-06, + "logits/chosen": -2.332122564315796, + "logits/rejected": -2.350968360900879, + "logps/chosen": -23.035541534423828, + "logps/rejected": -17.014999389648438, + "loss": 0.9551, + "rewards/accuracies": 0.71875, + "rewards/chosen": 0.036588456481695175, + "rewards/margins": 0.7014162540435791, + "rewards/rejected": -0.6648277640342712, + "step": 361 + }, + { + "epoch": 0.7947310647639956, + "grad_norm": 5.20027494430542, + "learning_rate": 3.816565310465919e-06, + "logits/chosen": -2.394535541534424, + "logits/rejected": -2.278174877166748, + "logps/chosen": -19.75065803527832, + "logps/rejected": -12.300567626953125, + "loss": 0.9152, + "rewards/accuracies": 0.625, + "rewards/chosen": -0.08690013736486435, + "rewards/margins": 0.452328085899353, + "rewards/rejected": -0.5392282605171204, + "step": 362 + }, + { + "epoch": 0.7969264544456641, + "grad_norm": 4.143304347991943, + "learning_rate": 3.8083126554982026e-06, + "logits/chosen": -2.3074147701263428, + "logits/rejected": -2.3378899097442627, + "logps/chosen": -20.852561950683594, + "logps/rejected": -10.050654411315918, + "loss": 0.8272, + "rewards/accuracies": 0.71875, + "rewards/chosen": -0.06546229869127274, + "rewards/margins": 0.758281409740448, + "rewards/rejected": -0.8237438201904297, + "step": 363 + }, + { + "epoch": 0.7991218441273326, + "grad_norm": 4.118852138519287, + "learning_rate": 3.8000403198230385e-06, + "logits/chosen": -2.343893051147461, + "logits/rejected": -2.3883817195892334, + "logps/chosen": -24.624427795410156, + "logps/rejected": -15.104917526245117, + "loss": 0.9192, + "rewards/accuracies": 0.65625, + "rewards/chosen": 0.12434723973274231, + "rewards/margins": 0.4829878509044647, + "rewards/rejected": -0.3586406111717224, + "step": 364 + }, + { + "epoch": 0.8013172338090011, + "grad_norm": 5.058257579803467, + "learning_rate": 3.7917484278796578e-06, + "logits/chosen": -2.4088287353515625, + "logits/rejected": -2.468463659286499, + "logps/chosen": -18.072254180908203, + "logps/rejected": -9.308141708374023, + "loss": 0.7411, + "rewards/accuracies": 0.78125, + "rewards/chosen": 0.022563787177205086, + "rewards/margins": 0.6568112373352051, + "rewards/rejected": -0.6342474222183228, + "step": 365 + }, + { + "epoch": 0.8035126234906695, + "grad_norm": 6.365854740142822, + "learning_rate": 3.7834371044014695e-06, + "logits/chosen": -2.3097360134124756, + "logits/rejected": -2.4446170330047607, + "logps/chosen": -17.25137710571289, + "logps/rejected": -13.248261451721191, + "loss": 0.8844, + "rewards/accuracies": 0.8125, + "rewards/chosen": 0.03468625992536545, + "rewards/margins": 0.5087876915931702, + "rewards/rejected": -0.4741014242172241, + "step": 366 + }, + { + "epoch": 0.8057080131723381, + "grad_norm": 4.799102783203125, + "learning_rate": 3.7751064744141886e-06, + "logits/chosen": -2.3456978797912598, + "logits/rejected": -2.2777140140533447, + "logps/chosen": -16.881250381469727, + "logps/rejected": -8.925056457519531, + "loss": 0.7494, + "rewards/accuracies": 0.84375, + "rewards/chosen": 0.0991620272397995, + "rewards/margins": 0.5507904887199402, + "rewards/rejected": -0.4516284465789795, + "step": 367 + }, + { + "epoch": 0.8079034028540066, + "grad_norm": 5.138694763183594, + "learning_rate": 3.7667566632339557e-06, + "logits/chosen": -2.25407338142395, + "logits/rejected": -2.3604111671447754, + "logps/chosen": -16.577442169189453, + "logps/rejected": -10.655071258544922, + "loss": 0.883, + "rewards/accuracies": 0.71875, + "rewards/chosen": 0.039467353373765945, + "rewards/margins": 0.4447363317012787, + "rewards/rejected": -0.40526896715164185, + "step": 368 + }, + { + "epoch": 0.8100987925356751, + "grad_norm": 6.1413397789001465, + "learning_rate": 3.75838779646545e-06, + "logits/chosen": -2.2719509601593018, + "logits/rejected": -2.270545721054077, + "logps/chosen": -16.89019203186035, + "logps/rejected": -10.996526718139648, + "loss": 0.8201, + "rewards/accuracies": 0.78125, + "rewards/chosen": 0.10040238499641418, + "rewards/margins": 0.5125969052314758, + "rewards/rejected": -0.41219455003738403, + "step": 369 + }, + { + "epoch": 0.8122941822173436, + "grad_norm": 4.1379499435424805, + "learning_rate": 3.7500000000000005e-06, + "logits/chosen": -2.3188323974609375, + "logits/rejected": -2.276256561279297, + "logps/chosen": -11.709390640258789, + "logps/rejected": -8.320552825927734, + "loss": 0.8101, + "rewards/accuracies": 0.6875, + "rewards/chosen": 0.06503432989120483, + "rewards/margins": 0.35664820671081543, + "rewards/rejected": -0.2916138768196106, + "step": 370 + }, + { + "epoch": 0.814489571899012, + "grad_norm": 4.974375247955322, + "learning_rate": 3.7415934000136907e-06, + "logits/chosen": -2.3040988445281982, + "logits/rejected": -2.378429412841797, + "logps/chosen": -16.86901092529297, + "logps/rejected": -10.178057670593262, + "loss": 0.7936, + "rewards/accuracies": 0.6875, + "rewards/chosen": 0.0352807454764843, + "rewards/margins": 0.6562831401824951, + "rewards/rejected": -0.6210023760795593, + "step": 371 + }, + { + "epoch": 0.8166849615806806, + "grad_norm": 5.170460224151611, + "learning_rate": 3.7331681229654637e-06, + "logits/chosen": -2.3687679767608643, + "logits/rejected": -2.406930923461914, + "logps/chosen": -19.379417419433594, + "logps/rejected": -12.089089393615723, + "loss": 0.8487, + "rewards/accuracies": 0.84375, + "rewards/chosen": 0.03307601809501648, + "rewards/margins": 0.6642818450927734, + "rewards/rejected": -0.6312057971954346, + "step": 372 + }, + { + "epoch": 0.818880351262349, + "grad_norm": 6.528963088989258, + "learning_rate": 3.724724295595218e-06, + "logits/chosen": -2.318371534347534, + "logits/rejected": -2.340467929840088, + "logps/chosen": -17.74517822265625, + "logps/rejected": -13.328447341918945, + "loss": 0.903, + "rewards/accuracies": 0.8125, + "rewards/chosen": 0.042347729206085205, + "rewards/margins": 0.4485569894313812, + "rewards/rejected": -0.4062092900276184, + "step": 373 + }, + { + "epoch": 0.8210757409440176, + "grad_norm": 3.9055440425872803, + "learning_rate": 3.7162620449219e-06, + "logits/chosen": -2.387302875518799, + "logits/rejected": -2.3294808864593506, + "logps/chosen": -15.384821891784668, + "logps/rejected": -9.518472671508789, + "loss": 0.802, + "rewards/accuracies": 0.65625, + "rewards/chosen": 0.06279882788658142, + "rewards/margins": 0.3927484154701233, + "rewards/rejected": -0.32994961738586426, + "step": 374 + }, + { + "epoch": 0.823271130625686, + "grad_norm": 4.305942058563232, + "learning_rate": 3.7077814982415966e-06, + "logits/chosen": -2.2834930419921875, + "logits/rejected": -2.1856932640075684, + "logps/chosen": -15.38658332824707, + "logps/rejected": -9.867746353149414, + "loss": 0.8521, + "rewards/accuracies": 0.8125, + "rewards/chosen": 0.12428806722164154, + "rewards/margins": 0.3732988238334656, + "rewards/rejected": -0.24901077151298523, + "step": 375 + }, + { + "epoch": 0.8254665203073546, + "grad_norm": 4.016692161560059, + "learning_rate": 3.699282783125616e-06, + "logits/chosen": -2.3968873023986816, + "logits/rejected": -2.3753557205200195, + "logps/chosen": -15.990100860595703, + "logps/rejected": -9.94045639038086, + "loss": 0.7915, + "rewards/accuracies": 0.8125, + "rewards/chosen": 0.0717865377664566, + "rewards/margins": 0.5052440166473389, + "rewards/rejected": -0.43345755338668823, + "step": 376 + }, + { + "epoch": 0.827661909989023, + "grad_norm": 5.630831241607666, + "learning_rate": 3.6907660274185723e-06, + "logits/chosen": -2.2964377403259277, + "logits/rejected": -2.360926389694214, + "logps/chosen": -20.974613189697266, + "logps/rejected": -11.94788646697998, + "loss": 0.8318, + "rewards/accuracies": 0.75, + "rewards/chosen": -0.12485136836767197, + "rewards/margins": 0.8055430054664612, + "rewards/rejected": -0.9303942918777466, + "step": 377 + }, + { + "epoch": 0.8298572996706916, + "grad_norm": 4.928332805633545, + "learning_rate": 3.6822313592364594e-06, + "logits/chosen": -2.358898401260376, + "logits/rejected": -2.3022220134735107, + "logps/chosen": -19.222505569458008, + "logps/rejected": -9.495677947998047, + "loss": 0.7366, + "rewards/accuracies": 0.75, + "rewards/chosen": 0.0382472425699234, + "rewards/margins": 0.7630552053451538, + "rewards/rejected": -0.7248079776763916, + "step": 378 + }, + { + "epoch": 0.8320526893523601, + "grad_norm": 11.10118293762207, + "learning_rate": 3.6736789069647273e-06, + "logits/chosen": -2.3210902214050293, + "logits/rejected": -2.2445425987243652, + "logps/chosen": -17.241561889648438, + "logps/rejected": -12.101102828979492, + "loss": 0.9336, + "rewards/accuracies": 0.65625, + "rewards/chosen": -0.012493027374148369, + "rewards/margins": 0.4085111916065216, + "rewards/rejected": -0.4210042655467987, + "step": 379 + }, + { + "epoch": 0.8342480790340285, + "grad_norm": 6.272946834564209, + "learning_rate": 3.665108799256348e-06, + "logits/chosen": -2.1888327598571777, + "logits/rejected": -2.325762987136841, + "logps/chosen": -20.95864486694336, + "logps/rejected": -8.965181350708008, + "loss": 0.7478, + "rewards/accuracies": 0.8125, + "rewards/chosen": 0.08120650053024292, + "rewards/margins": 0.6599326133728027, + "rewards/rejected": -0.5787261128425598, + "step": 380 + }, + { + "epoch": 0.8364434687156971, + "grad_norm": 5.793782711029053, + "learning_rate": 3.6565211650298787e-06, + "logits/chosen": -2.251807689666748, + "logits/rejected": -2.2617058753967285, + "logps/chosen": -28.702529907226562, + "logps/rejected": -15.13422966003418, + "loss": 0.8766, + "rewards/accuracies": 0.75, + "rewards/chosen": 0.003576405346393585, + "rewards/margins": 0.8133214116096497, + "rewards/rejected": -0.8097449541091919, + "step": 381 + }, + { + "epoch": 0.8386388583973655, + "grad_norm": 6.972100734710693, + "learning_rate": 3.6479161334675294e-06, + "logits/chosen": -2.285806179046631, + "logits/rejected": -2.4609298706054688, + "logps/chosen": -18.277084350585938, + "logps/rejected": -12.115345001220703, + "loss": 0.8225, + "rewards/accuracies": 0.84375, + "rewards/chosen": 0.0019068922847509384, + "rewards/margins": 0.6515531539916992, + "rewards/rejected": -0.6496463418006897, + "step": 382 + }, + { + "epoch": 0.8408342480790341, + "grad_norm": 4.72037935256958, + "learning_rate": 3.639293834013211e-06, + "logits/chosen": -2.3857169151306152, + "logits/rejected": -2.2780425548553467, + "logps/chosen": -23.369903564453125, + "logps/rejected": -15.874404907226562, + "loss": 0.9017, + "rewards/accuracies": 0.71875, + "rewards/chosen": 0.0180068202316761, + "rewards/margins": 0.7297979593276978, + "rewards/rejected": -0.7117910385131836, + "step": 383 + }, + { + "epoch": 0.8430296377607025, + "grad_norm": 6.700748443603516, + "learning_rate": 3.6306543963705943e-06, + "logits/chosen": -2.333336591720581, + "logits/rejected": -2.3500876426696777, + "logps/chosen": -20.27480125427246, + "logps/rejected": -11.234090805053711, + "loss": 0.7802, + "rewards/accuracies": 0.78125, + "rewards/chosen": 0.02051311358809471, + "rewards/margins": 0.6360718011856079, + "rewards/rejected": -0.6155586242675781, + "step": 384 + }, + { + "epoch": 0.845225027442371, + "grad_norm": 4.831250190734863, + "learning_rate": 3.621997950501156e-06, + "logits/chosen": -2.2861812114715576, + "logits/rejected": -2.3506617546081543, + "logps/chosen": -18.32001304626465, + "logps/rejected": -11.220377922058105, + "loss": 0.8039, + "rewards/accuracies": 0.84375, + "rewards/chosen": 0.03276555612683296, + "rewards/margins": 0.6321474313735962, + "rewards/rejected": -0.5993818640708923, + "step": 385 + }, + { + "epoch": 0.8474204171240395, + "grad_norm": 7.575469017028809, + "learning_rate": 3.613324626622224e-06, + "logits/chosen": -2.3208696842193604, + "logits/rejected": -2.335655689239502, + "logps/chosen": -16.124113082885742, + "logps/rejected": -13.462167739868164, + "loss": 1.0063, + "rewards/accuracies": 0.6875, + "rewards/chosen": -0.0590025819838047, + "rewards/margins": 0.3688778877258301, + "rewards/rejected": -0.4278804659843445, + "step": 386 + }, + { + "epoch": 0.849615806805708, + "grad_norm": 9.63671875, + "learning_rate": 3.60463455520502e-06, + "logits/chosen": -2.388012647628784, + "logits/rejected": -2.3075461387634277, + "logps/chosen": -18.78927230834961, + "logps/rejected": -14.575672149658203, + "loss": 0.9509, + "rewards/accuracies": 0.65625, + "rewards/chosen": -0.1962830275297165, + "rewards/margins": 0.41697803139686584, + "rewards/rejected": -0.6132611036300659, + "step": 387 + }, + { + "epoch": 0.8518111964873765, + "grad_norm": 5.019967079162598, + "learning_rate": 3.595927866972694e-06, + "logits/chosen": -2.301435708999634, + "logits/rejected": -2.2961511611938477, + "logps/chosen": -20.059595108032227, + "logps/rejected": -13.45213794708252, + "loss": 0.916, + "rewards/accuracies": 0.6875, + "rewards/chosen": -0.027182452380657196, + "rewards/margins": 0.6479660868644714, + "rewards/rejected": -0.6751485466957092, + "step": 388 + }, + { + "epoch": 0.854006586169045, + "grad_norm": 7.656889915466309, + "learning_rate": 3.587204692898363e-06, + "logits/chosen": -2.2869014739990234, + "logits/rejected": -2.3017992973327637, + "logps/chosen": -20.62851905822754, + "logps/rejected": -9.424978256225586, + "loss": 0.7464, + "rewards/accuracies": 0.78125, + "rewards/chosen": 0.03314004838466644, + "rewards/margins": 0.8571538925170898, + "rewards/rejected": -0.8240138292312622, + "step": 389 + }, + { + "epoch": 0.8562019758507134, + "grad_norm": 5.941650390625, + "learning_rate": 3.578465164203134e-06, + "logits/chosen": -2.2174339294433594, + "logits/rejected": -2.256338357925415, + "logps/chosen": -17.569561004638672, + "logps/rejected": -12.100666046142578, + "loss": 0.8744, + "rewards/accuracies": 0.75, + "rewards/chosen": 0.05165117233991623, + "rewards/margins": 0.6384047269821167, + "rewards/rejected": -0.5867536067962646, + "step": 390 + }, + { + "epoch": 0.858397365532382, + "grad_norm": 4.409201622009277, + "learning_rate": 3.5697094123541357e-06, + "logits/chosen": -2.264608860015869, + "logits/rejected": -2.2907967567443848, + "logps/chosen": -19.413509368896484, + "logps/rejected": -10.549042701721191, + "loss": 0.8083, + "rewards/accuracies": 0.6875, + "rewards/chosen": -0.005588045343756676, + "rewards/margins": 0.5789726972579956, + "rewards/rejected": -0.584560751914978, + "step": 391 + }, + { + "epoch": 0.8605927552140505, + "grad_norm": 5.469266891479492, + "learning_rate": 3.5609375690625384e-06, + "logits/chosen": -2.2814698219299316, + "logits/rejected": -2.2320809364318848, + "logps/chosen": -20.82750701904297, + "logps/rejected": -10.755342483520508, + "loss": 0.9039, + "rewards/accuracies": 0.71875, + "rewards/chosen": -0.09153535217046738, + "rewards/margins": 0.5601884126663208, + "rewards/rejected": -0.6517237424850464, + "step": 392 + }, + { + "epoch": 0.862788144895719, + "grad_norm": 4.7851176261901855, + "learning_rate": 3.552149766281573e-06, + "logits/chosen": -2.3673245906829834, + "logits/rejected": -2.288597583770752, + "logps/chosen": -17.121660232543945, + "logps/rejected": -10.34488296508789, + "loss": 0.7823, + "rewards/accuracies": 0.78125, + "rewards/chosen": 0.08655978739261627, + "rewards/margins": 0.6128506660461426, + "rewards/rejected": -0.5262908935546875, + "step": 393 + }, + { + "epoch": 0.8649835345773875, + "grad_norm": 4.078333854675293, + "learning_rate": 3.543346136204545e-06, + "logits/chosen": -2.2628226280212402, + "logits/rejected": -2.4437499046325684, + "logps/chosen": -21.973432540893555, + "logps/rejected": -11.049983024597168, + "loss": 0.7704, + "rewards/accuracies": 0.875, + "rewards/chosen": 0.13885267078876495, + "rewards/margins": 0.9487197995185852, + "rewards/rejected": -0.8098670840263367, + "step": 394 + }, + { + "epoch": 0.867178924259056, + "grad_norm": 5.391472339630127, + "learning_rate": 3.5345268112628485e-06, + "logits/chosen": -2.2225139141082764, + "logits/rejected": -2.41272234916687, + "logps/chosen": -20.41250991821289, + "logps/rejected": -9.64039421081543, + "loss": 0.7121, + "rewards/accuracies": 0.78125, + "rewards/chosen": 0.0024206284433603287, + "rewards/margins": 0.9180695414543152, + "rewards/rejected": -0.915648877620697, + "step": 395 + }, + { + "epoch": 0.8693743139407245, + "grad_norm": 4.787206172943115, + "learning_rate": 3.5256919241239712e-06, + "logits/chosen": -2.36836576461792, + "logits/rejected": -2.396129608154297, + "logps/chosen": -20.147674560546875, + "logps/rejected": -12.847021102905273, + "loss": 0.865, + "rewards/accuracies": 0.84375, + "rewards/chosen": -0.02502620406448841, + "rewards/margins": 0.755872368812561, + "rewards/rejected": -0.7808985114097595, + "step": 396 + }, + { + "epoch": 0.8715697036223929, + "grad_norm": 5.62094259262085, + "learning_rate": 3.516841607689501e-06, + "logits/chosen": -2.3200628757476807, + "logits/rejected": -2.369778633117676, + "logps/chosen": -21.13600730895996, + "logps/rejected": -13.838922500610352, + "loss": 0.8936, + "rewards/accuracies": 0.75, + "rewards/chosen": -0.09514954686164856, + "rewards/margins": 0.6872110962867737, + "rewards/rejected": -0.7823606729507446, + "step": 397 + }, + { + "epoch": 0.8737650933040615, + "grad_norm": 6.053920269012451, + "learning_rate": 3.5079759950931257e-06, + "logits/chosen": -2.302199363708496, + "logits/rejected": -2.4088432788848877, + "logps/chosen": -19.452991485595703, + "logps/rejected": -10.391996383666992, + "loss": 0.7719, + "rewards/accuracies": 0.75, + "rewards/chosen": -0.04446466267108917, + "rewards/margins": 0.6437113881111145, + "rewards/rejected": -0.6881760358810425, + "step": 398 + }, + { + "epoch": 0.8759604829857299, + "grad_norm": 5.638072967529297, + "learning_rate": 3.499095219698631e-06, + "logits/chosen": -2.2861905097961426, + "logits/rejected": -2.3225479125976562, + "logps/chosen": -19.642776489257812, + "logps/rejected": -10.90954875946045, + "loss": 0.7718, + "rewards/accuracies": 0.875, + "rewards/chosen": 0.08755317330360413, + "rewards/margins": 0.6167296171188354, + "rewards/rejected": -0.5291764736175537, + "step": 399 + }, + { + "epoch": 0.8781558726673985, + "grad_norm": 6.903406620025635, + "learning_rate": 3.4901994150978926e-06, + "logits/chosen": -2.307342529296875, + "logits/rejected": -2.337017059326172, + "logps/chosen": -19.130626678466797, + "logps/rejected": -10.1241455078125, + "loss": 0.8591, + "rewards/accuracies": 0.65625, + "rewards/chosen": -0.03486700728535652, + "rewards/margins": 0.5795115232467651, + "rewards/rejected": -0.6143784523010254, + "step": 400 + }, + { + "epoch": 0.8803512623490669, + "grad_norm": 4.02300500869751, + "learning_rate": 3.481288715108868e-06, + "logits/chosen": -2.3123385906219482, + "logits/rejected": -2.2987749576568604, + "logps/chosen": -20.062665939331055, + "logps/rejected": -11.380792617797852, + "loss": 0.8085, + "rewards/accuracies": 0.75, + "rewards/chosen": 0.026614921167492867, + "rewards/margins": 0.7673150300979614, + "rewards/rejected": -0.7407000064849854, + "step": 401 + }, + { + "epoch": 0.8825466520307355, + "grad_norm": 3.974309206008911, + "learning_rate": 3.4723632537735846e-06, + "logits/chosen": -2.3131566047668457, + "logits/rejected": -2.3440208435058594, + "logps/chosen": -16.116439819335938, + "logps/rejected": -11.196529388427734, + "loss": 0.8536, + "rewards/accuracies": 0.75, + "rewards/chosen": 0.033491406589746475, + "rewards/margins": 0.460104376077652, + "rewards/rejected": -0.4266130030155182, + "step": 402 + }, + { + "epoch": 0.884742041712404, + "grad_norm": 9.103083610534668, + "learning_rate": 3.4634231653561213e-06, + "logits/chosen": -2.26672625541687, + "logits/rejected": -2.3151166439056396, + "logps/chosen": -18.1038875579834, + "logps/rejected": -10.783166885375977, + "loss": 0.9248, + "rewards/accuracies": 0.5, + "rewards/chosen": -0.09668270498514175, + "rewards/margins": 0.36572739481925964, + "rewards/rejected": -0.4624100923538208, + "step": 403 + }, + { + "epoch": 0.8869374313940724, + "grad_norm": 8.053287506103516, + "learning_rate": 3.454468584340588e-06, + "logits/chosen": -2.4167327880859375, + "logits/rejected": -2.391662120819092, + "logps/chosen": -16.634321212768555, + "logps/rejected": -11.324498176574707, + "loss": 0.9218, + "rewards/accuracies": 0.59375, + "rewards/chosen": -0.11378289759159088, + "rewards/margins": 0.2954775393009186, + "rewards/rejected": -0.40926042199134827, + "step": 404 + }, + { + "epoch": 0.889132821075741, + "grad_norm": 10.441896438598633, + "learning_rate": 3.4454996454291066e-06, + "logits/chosen": -2.254704236984253, + "logits/rejected": -2.3380165100097656, + "logps/chosen": -19.833484649658203, + "logps/rejected": -12.246344566345215, + "loss": 0.9228, + "rewards/accuracies": 0.625, + "rewards/chosen": -0.06332497298717499, + "rewards/margins": 0.4893012046813965, + "rewards/rejected": -0.5526261329650879, + "step": 405 + }, + { + "epoch": 0.8913282107574094, + "grad_norm": 6.624183654785156, + "learning_rate": 3.436516483539781e-06, + "logits/chosen": -2.284971237182617, + "logits/rejected": -2.3368124961853027, + "logps/chosen": -16.181856155395508, + "logps/rejected": -10.476940155029297, + "loss": 0.8555, + "rewards/accuracies": 0.65625, + "rewards/chosen": 0.09641481190919876, + "rewards/margins": 0.5150280594825745, + "rewards/rejected": -0.4186131954193115, + "step": 406 + }, + { + "epoch": 0.893523600439078, + "grad_norm": 15.161913871765137, + "learning_rate": 3.4275192338046677e-06, + "logits/chosen": -2.441598415374756, + "logits/rejected": -2.354428768157959, + "logps/chosen": -15.579620361328125, + "logps/rejected": -15.171042442321777, + "loss": 1.1264, + "rewards/accuracies": 0.59375, + "rewards/chosen": -0.20871956646442413, + "rewards/margins": 0.14662565290927887, + "rewards/rejected": -0.355345219373703, + "step": 407 + }, + { + "epoch": 0.8957189901207464, + "grad_norm": 4.21067476272583, + "learning_rate": 3.4185080315677454e-06, + "logits/chosen": -2.347850799560547, + "logits/rejected": -2.31205677986145, + "logps/chosen": -21.443716049194336, + "logps/rejected": -13.075089454650879, + "loss": 0.7881, + "rewards/accuracies": 0.90625, + "rewards/chosen": 0.07311268150806427, + "rewards/margins": 0.9533619284629822, + "rewards/rejected": -0.8802492618560791, + "step": 408 + }, + { + "epoch": 0.897914379802415, + "grad_norm": 7.904491424560547, + "learning_rate": 3.409483012382879e-06, + "logits/chosen": -2.24735689163208, + "logits/rejected": -2.2707090377807617, + "logps/chosen": -22.61405372619629, + "logps/rejected": -13.15417766571045, + "loss": 0.8115, + "rewards/accuracies": 0.75, + "rewards/chosen": -0.1565491259098053, + "rewards/margins": 0.9658958911895752, + "rewards/rejected": -1.122444987297058, + "step": 409 + }, + { + "epoch": 0.9001097694840834, + "grad_norm": 6.310301303863525, + "learning_rate": 3.400444312011776e-06, + "logits/chosen": -2.3683767318725586, + "logits/rejected": -2.3440041542053223, + "logps/chosen": -17.640872955322266, + "logps/rejected": -10.977629661560059, + "loss": 0.8158, + "rewards/accuracies": 0.6875, + "rewards/chosen": 0.028235310688614845, + "rewards/margins": 0.590648353099823, + "rewards/rejected": -0.5624130368232727, + "step": 410 + }, + { + "epoch": 0.9023051591657519, + "grad_norm": 5.326711654663086, + "learning_rate": 3.3913920664219504e-06, + "logits/chosen": -2.447878837585449, + "logits/rejected": -2.4349117279052734, + "logps/chosen": -19.840436935424805, + "logps/rejected": -8.293453216552734, + "loss": 0.7169, + "rewards/accuracies": 0.75, + "rewards/chosen": 0.06985548138618469, + "rewards/margins": 0.8774117231369019, + "rewards/rejected": -0.8075562715530396, + "step": 411 + }, + { + "epoch": 0.9045005488474204, + "grad_norm": 14.260117530822754, + "learning_rate": 3.3823264117846722e-06, + "logits/chosen": -2.368217945098877, + "logits/rejected": -2.411201238632202, + "logps/chosen": -18.50075912475586, + "logps/rejected": -12.620443344116211, + "loss": 0.9106, + "rewards/accuracies": 0.625, + "rewards/chosen": -0.04634511470794678, + "rewards/margins": 0.660761833190918, + "rewards/rejected": -0.7071069478988647, + "step": 412 + }, + { + "epoch": 0.9066959385290889, + "grad_norm": 6.360561847686768, + "learning_rate": 3.3732474844729235e-06, + "logits/chosen": -2.2475552558898926, + "logits/rejected": -2.3121492862701416, + "logps/chosen": -19.70197296142578, + "logps/rejected": -11.163705825805664, + "loss": 0.8478, + "rewards/accuracies": 0.71875, + "rewards/chosen": -0.06956078857183456, + "rewards/margins": 0.6349106431007385, + "rewards/rejected": -0.7044714093208313, + "step": 413 + }, + { + "epoch": 0.9088913282107574, + "grad_norm": 6.33591365814209, + "learning_rate": 3.3641554210593417e-06, + "logits/chosen": -2.2870852947235107, + "logits/rejected": -2.4032349586486816, + "logps/chosen": -21.093215942382812, + "logps/rejected": -10.273629188537598, + "loss": 0.7941, + "rewards/accuracies": 0.78125, + "rewards/chosen": -0.019855815917253494, + "rewards/margins": 0.6967600584030151, + "rewards/rejected": -0.7166157960891724, + "step": 414 + }, + { + "epoch": 0.9110867178924259, + "grad_norm": 6.404439926147461, + "learning_rate": 3.3550503583141726e-06, + "logits/chosen": -2.3385732173919678, + "logits/rejected": -2.3589258193969727, + "logps/chosen": -20.491695404052734, + "logps/rejected": -14.065013885498047, + "loss": 0.9211, + "rewards/accuracies": 0.6875, + "rewards/chosen": -0.04469328746199608, + "rewards/margins": 0.5574424862861633, + "rewards/rejected": -0.6021357774734497, + "step": 415 + }, + { + "epoch": 0.9132821075740944, + "grad_norm": 5.92446231842041, + "learning_rate": 3.3459324332032035e-06, + "logits/chosen": -2.390655279159546, + "logits/rejected": -2.313477039337158, + "logps/chosen": -20.9737548828125, + "logps/rejected": -11.82977294921875, + "loss": 0.791, + "rewards/accuracies": 0.65625, + "rewards/chosen": -0.037018876522779465, + "rewards/margins": 0.7438760995864868, + "rewards/rejected": -0.7808948755264282, + "step": 416 + }, + { + "epoch": 0.9154774972557629, + "grad_norm": 7.234256267547607, + "learning_rate": 3.3368017828857117e-06, + "logits/chosen": -2.3002657890319824, + "logits/rejected": -2.3773255348205566, + "logps/chosen": -26.463184356689453, + "logps/rejected": -14.281316757202148, + "loss": 0.8885, + "rewards/accuracies": 0.78125, + "rewards/chosen": -0.0562882199883461, + "rewards/margins": 0.8361290693283081, + "rewards/rejected": -0.8924172520637512, + "step": 417 + }, + { + "epoch": 0.9176728869374314, + "grad_norm": 8.097594261169434, + "learning_rate": 3.3276585447123957e-06, + "logits/chosen": -2.2877120971679688, + "logits/rejected": -2.381965398788452, + "logps/chosen": -18.890178680419922, + "logps/rejected": -15.372221946716309, + "loss": 1.0001, + "rewards/accuracies": 0.75, + "rewards/chosen": -0.12048324942588806, + "rewards/margins": 0.5791645646095276, + "rewards/rejected": -0.6996477842330933, + "step": 418 + }, + { + "epoch": 0.9198682766190999, + "grad_norm": 5.180337905883789, + "learning_rate": 3.318502856223311e-06, + "logits/chosen": -2.308959484100342, + "logits/rejected": -2.3321940898895264, + "logps/chosen": -22.121414184570312, + "logps/rejected": -15.669811248779297, + "loss": 0.9739, + "rewards/accuracies": 0.65625, + "rewards/chosen": -0.04433257132768631, + "rewards/margins": 0.5762405395507812, + "rewards/rejected": -0.620573103427887, + "step": 419 + }, + { + "epoch": 0.9220636663007684, + "grad_norm": 5.768693447113037, + "learning_rate": 3.3093348551458033e-06, + "logits/chosen": -2.293351411819458, + "logits/rejected": -2.3098649978637695, + "logps/chosen": -19.407869338989258, + "logps/rejected": -13.331077575683594, + "loss": 0.9355, + "rewards/accuracies": 0.75, + "rewards/chosen": -0.013614770025014877, + "rewards/margins": 0.37932369112968445, + "rewards/rejected": -0.3929384648799896, + "step": 420 + }, + { + "epoch": 0.9242590559824369, + "grad_norm": 5.084871292114258, + "learning_rate": 3.300154679392429e-06, + "logits/chosen": -2.372471809387207, + "logits/rejected": -2.372732162475586, + "logps/chosen": -18.58143424987793, + "logps/rejected": -10.904744148254395, + "loss": 0.8029, + "rewards/accuracies": 0.78125, + "rewards/chosen": 0.06652218848466873, + "rewards/margins": 0.5077332854270935, + "rewards/rejected": -0.4412110447883606, + "step": 421 + }, + { + "epoch": 0.9264544456641054, + "grad_norm": 5.599188327789307, + "learning_rate": 3.2909624670588915e-06, + "logits/chosen": -2.2680983543395996, + "logits/rejected": -2.403550386428833, + "logps/chosen": -22.577377319335938, + "logps/rejected": -16.19898223876953, + "loss": 0.9287, + "rewards/accuracies": 0.8125, + "rewards/chosen": -0.008984517306089401, + "rewards/margins": 0.7310069799423218, + "rewards/rejected": -0.7399914264678955, + "step": 422 + }, + { + "epoch": 0.9286498353457738, + "grad_norm": 6.535630226135254, + "learning_rate": 3.281758356421955e-06, + "logits/chosen": -2.2871334552764893, + "logits/rejected": -2.379993438720703, + "logps/chosen": -23.122037887573242, + "logps/rejected": -11.495004653930664, + "loss": 0.7905, + "rewards/accuracies": 0.78125, + "rewards/chosen": 0.10842779278755188, + "rewards/margins": 0.7593700289726257, + "rewards/rejected": -0.6509422063827515, + "step": 423 + }, + { + "epoch": 0.9308452250274424, + "grad_norm": 4.752484321594238, + "learning_rate": 3.272542485937369e-06, + "logits/chosen": -2.317404270172119, + "logits/rejected": -2.3489813804626465, + "logps/chosen": -19.68863296508789, + "logps/rejected": -11.732726097106934, + "loss": 0.8028, + "rewards/accuracies": 0.78125, + "rewards/chosen": 0.04726914316415787, + "rewards/margins": 0.7166731953620911, + "rewards/rejected": -0.6694040298461914, + "step": 424 + }, + { + "epoch": 0.9330406147091108, + "grad_norm": 5.882680416107178, + "learning_rate": 3.2633149942377835e-06, + "logits/chosen": -2.2992334365844727, + "logits/rejected": -2.4041848182678223, + "logps/chosen": -19.215017318725586, + "logps/rejected": -13.420120239257812, + "loss": 0.8508, + "rewards/accuracies": 0.65625, + "rewards/chosen": 0.1038583368062973, + "rewards/margins": 0.7461327910423279, + "rewards/rejected": -0.6422744393348694, + "step": 425 + }, + { + "epoch": 0.9352360043907794, + "grad_norm": 7.522842884063721, + "learning_rate": 3.2540760201306643e-06, + "logits/chosen": -2.331803798675537, + "logits/rejected": -2.3895986080169678, + "logps/chosen": -21.03508758544922, + "logps/rejected": -12.05073356628418, + "loss": 0.7629, + "rewards/accuracies": 0.84375, + "rewards/chosen": 0.06679324805736542, + "rewards/margins": 0.789514422416687, + "rewards/rejected": -0.7227212190628052, + "step": 426 + }, + { + "epoch": 0.9374313940724479, + "grad_norm": 11.598859786987305, + "learning_rate": 3.244825702596205e-06, + "logits/chosen": -2.3380470275878906, + "logits/rejected": -2.36613392829895, + "logps/chosen": -16.171180725097656, + "logps/rejected": -13.282466888427734, + "loss": 1.0535, + "rewards/accuracies": 0.65625, + "rewards/chosen": -0.14831915497779846, + "rewards/margins": 0.219946950674057, + "rewards/rejected": -0.36826610565185547, + "step": 427 + }, + { + "epoch": 0.9396267837541163, + "grad_norm": 5.60055685043335, + "learning_rate": 3.2355641807852377e-06, + "logits/chosen": -2.3118081092834473, + "logits/rejected": -2.328697681427002, + "logps/chosen": -21.737201690673828, + "logps/rejected": -8.444561958312988, + "loss": 0.6397, + "rewards/accuracies": 0.875, + "rewards/chosen": 0.16470521688461304, + "rewards/margins": 1.143484115600586, + "rewards/rejected": -0.9787789583206177, + "step": 428 + }, + { + "epoch": 0.9418221734357849, + "grad_norm": 7.712846279144287, + "learning_rate": 3.2262915940171373e-06, + "logits/chosen": -2.3408477306365967, + "logits/rejected": -2.258118152618408, + "logps/chosen": -18.901592254638672, + "logps/rejected": -9.10704231262207, + "loss": 0.7194, + "rewards/accuracies": 0.84375, + "rewards/chosen": 0.12541034817695618, + "rewards/margins": 0.7345283031463623, + "rewards/rejected": -0.6091179847717285, + "step": 429 + }, + { + "epoch": 0.9440175631174533, + "grad_norm": 5.265126705169678, + "learning_rate": 3.217008081777726e-06, + "logits/chosen": -2.3288192749023438, + "logits/rejected": -2.3424811363220215, + "logps/chosen": -17.280908584594727, + "logps/rejected": -12.148330688476562, + "loss": 0.8629, + "rewards/accuracies": 0.75, + "rewards/chosen": 0.03913713991641998, + "rewards/margins": 0.5172592401504517, + "rewards/rejected": -0.47812211513519287, + "step": 430 + }, + { + "epoch": 0.9462129527991219, + "grad_norm": 6.47225284576416, + "learning_rate": 3.2077137837171764e-06, + "logits/chosen": -2.3356518745422363, + "logits/rejected": -2.3391013145446777, + "logps/chosen": -24.702030181884766, + "logps/rejected": -14.061824798583984, + "loss": 0.8684, + "rewards/accuracies": 0.65625, + "rewards/chosen": 0.02924323081970215, + "rewards/margins": 0.9203999042510986, + "rewards/rejected": -0.8911566734313965, + "step": 431 + }, + { + "epoch": 0.9484083424807903, + "grad_norm": 7.641210556030273, + "learning_rate": 3.1984088396479113e-06, + "logits/chosen": -2.343367338180542, + "logits/rejected": -2.301558494567871, + "logps/chosen": -18.52604103088379, + "logps/rejected": -10.527841567993164, + "loss": 0.8235, + "rewards/accuracies": 0.78125, + "rewards/chosen": 0.03001457452774048, + "rewards/margins": 0.6643778085708618, + "rewards/rejected": -0.6343631744384766, + "step": 432 + }, + { + "epoch": 0.9506037321624589, + "grad_norm": 5.307621955871582, + "learning_rate": 3.189093389542498e-06, + "logits/chosen": -2.230017900466919, + "logits/rejected": -2.2622296810150146, + "logps/chosen": -15.772552490234375, + "logps/rejected": -10.832294464111328, + "loss": 0.9047, + "rewards/accuracies": 0.65625, + "rewards/chosen": 0.02474249340593815, + "rewards/margins": 0.5498454570770264, + "rewards/rejected": -0.525102972984314, + "step": 433 + }, + { + "epoch": 0.9527991218441273, + "grad_norm": 8.881011962890625, + "learning_rate": 3.179767573531546e-06, + "logits/chosen": -2.3451647758483887, + "logits/rejected": -2.2926852703094482, + "logps/chosen": -16.5456600189209, + "logps/rejected": -11.103011131286621, + "loss": 0.8291, + "rewards/accuracies": 0.6875, + "rewards/chosen": 0.05186466500163078, + "rewards/margins": 0.5443666577339172, + "rewards/rejected": -0.49250200390815735, + "step": 434 + }, + { + "epoch": 0.9549945115257958, + "grad_norm": 5.801877498626709, + "learning_rate": 3.1704315319015936e-06, + "logits/chosen": -2.3372323513031006, + "logits/rejected": -2.3091893196105957, + "logps/chosen": -23.148170471191406, + "logps/rejected": -16.703672409057617, + "loss": 0.8901, + "rewards/accuracies": 0.78125, + "rewards/chosen": -0.027854733169078827, + "rewards/margins": 0.7050666809082031, + "rewards/rejected": -0.7329213619232178, + "step": 435 + }, + { + "epoch": 0.9571899012074643, + "grad_norm": 5.348484039306641, + "learning_rate": 3.1610854050930063e-06, + "logits/chosen": -2.368006944656372, + "logits/rejected": -2.418947458267212, + "logps/chosen": -18.732362747192383, + "logps/rejected": -9.964558601379395, + "loss": 0.7543, + "rewards/accuracies": 0.8125, + "rewards/chosen": 0.043328557163476944, + "rewards/margins": 0.7478787302970886, + "rewards/rejected": -0.7045502066612244, + "step": 436 + }, + { + "epoch": 0.9593852908891328, + "grad_norm": 9.07236099243164, + "learning_rate": 3.1517293336978538e-06, + "logits/chosen": -2.2696292400360107, + "logits/rejected": -2.3099236488342285, + "logps/chosen": -15.583741188049316, + "logps/rejected": -17.732797622680664, + "loss": 1.1379, + "rewards/accuracies": 0.71875, + "rewards/chosen": -0.1528353989124298, + "rewards/margins": 0.2265656292438507, + "rewards/rejected": -0.3794010579586029, + "step": 437 + }, + { + "epoch": 0.9615806805708014, + "grad_norm": 6.247621536254883, + "learning_rate": 3.1423634584578056e-06, + "logits/chosen": -2.392490863800049, + "logits/rejected": -2.3658711910247803, + "logps/chosen": -18.44986343383789, + "logps/rejected": -14.234761238098145, + "loss": 0.8623, + "rewards/accuracies": 0.8125, + "rewards/chosen": 0.04860144108533859, + "rewards/margins": 0.7451760768890381, + "rewards/rejected": -0.6965745687484741, + "step": 438 + }, + { + "epoch": 0.9637760702524698, + "grad_norm": 5.417273998260498, + "learning_rate": 3.132987920262005e-06, + "logits/chosen": -2.3367209434509277, + "logits/rejected": -2.4058802127838135, + "logps/chosen": -17.406082153320312, + "logps/rejected": -8.535592079162598, + "loss": 0.7611, + "rewards/accuracies": 0.6875, + "rewards/chosen": 0.14041852951049805, + "rewards/margins": 0.6607614755630493, + "rewards/rejected": -0.5203429460525513, + "step": 439 + }, + { + "epoch": 0.9659714599341384, + "grad_norm": 7.8660359382629395, + "learning_rate": 3.1236028601449534e-06, + "logits/chosen": -2.2924911975860596, + "logits/rejected": -2.309866189956665, + "logps/chosen": -25.160966873168945, + "logps/rejected": -11.313871383666992, + "loss": 0.764, + "rewards/accuracies": 0.78125, + "rewards/chosen": 0.008873052895069122, + "rewards/margins": 1.0792553424835205, + "rewards/rejected": -1.0703822374343872, + "step": 440 + }, + { + "epoch": 0.9681668496158068, + "grad_norm": 7.840826034545898, + "learning_rate": 3.114208419284391e-06, + "logits/chosen": -2.3387792110443115, + "logits/rejected": -2.3440520763397217, + "logps/chosen": -18.8978271484375, + "logps/rejected": -12.489431381225586, + "loss": 0.8975, + "rewards/accuracies": 0.75, + "rewards/chosen": -0.06741246581077576, + "rewards/margins": 0.6665923595428467, + "rewards/rejected": -0.7340048551559448, + "step": 441 + }, + { + "epoch": 0.9703622392974753, + "grad_norm": 6.152930736541748, + "learning_rate": 3.1048047389991693e-06, + "logits/chosen": -2.3817009925842285, + "logits/rejected": -2.357701301574707, + "logps/chosen": -16.76677894592285, + "logps/rejected": -11.620492935180664, + "loss": 0.8242, + "rewards/accuracies": 0.75, + "rewards/chosen": 0.0717247724533081, + "rewards/margins": 0.5190142393112183, + "rewards/rejected": -0.44728946685791016, + "step": 442 + }, + { + "epoch": 0.9725576289791438, + "grad_norm": 5.161471366882324, + "learning_rate": 3.0953919607471276e-06, + "logits/chosen": -2.2455554008483887, + "logits/rejected": -2.2900137901306152, + "logps/chosen": -18.165666580200195, + "logps/rejected": -11.611608505249023, + "loss": 0.8377, + "rewards/accuracies": 0.78125, + "rewards/chosen": 0.08798500150442123, + "rewards/margins": 0.5571097135543823, + "rewards/rejected": -0.4691247344017029, + "step": 443 + }, + { + "epoch": 0.9747530186608123, + "grad_norm": 4.925886154174805, + "learning_rate": 3.0859702261229616e-06, + "logits/chosen": -2.255220651626587, + "logits/rejected": -2.2815744876861572, + "logps/chosen": -17.18117332458496, + "logps/rejected": -8.485081672668457, + "loss": 0.7083, + "rewards/accuracies": 0.8125, + "rewards/chosen": 0.16568660736083984, + "rewards/margins": 0.7109737396240234, + "rewards/rejected": -0.5452871322631836, + "step": 444 + }, + { + "epoch": 0.9769484083424808, + "grad_norm": 8.389778137207031, + "learning_rate": 3.0765396768561005e-06, + "logits/chosen": -2.2690978050231934, + "logits/rejected": -2.2897820472717285, + "logps/chosen": -16.80807876586914, + "logps/rejected": -12.906442642211914, + "loss": 0.9586, + "rewards/accuracies": 0.71875, + "rewards/chosen": 0.05918867886066437, + "rewards/margins": 0.3266395628452301, + "rewards/rejected": -0.26745086908340454, + "step": 445 + }, + { + "epoch": 0.9791437980241493, + "grad_norm": 9.920949935913086, + "learning_rate": 3.0671004548085675e-06, + "logits/chosen": -2.3090577125549316, + "logits/rejected": -2.2865028381347656, + "logps/chosen": -20.217845916748047, + "logps/rejected": -16.117212295532227, + "loss": 0.8778, + "rewards/accuracies": 0.8125, + "rewards/chosen": 0.07724549621343613, + "rewards/margins": 0.706253170967102, + "rewards/rejected": -0.6290077567100525, + "step": 446 + }, + { + "epoch": 0.9813391877058177, + "grad_norm": 8.199292182922363, + "learning_rate": 3.0576527019728485e-06, + "logits/chosen": -2.2236199378967285, + "logits/rejected": -2.326417922973633, + "logps/chosen": -24.821739196777344, + "logps/rejected": -13.166486740112305, + "loss": 0.9256, + "rewards/accuracies": 0.71875, + "rewards/chosen": -0.04876386374235153, + "rewards/margins": 0.7376790046691895, + "rewards/rejected": -0.7864428758621216, + "step": 447 + }, + { + "epoch": 0.9835345773874863, + "grad_norm": 3.362391710281372, + "learning_rate": 3.0481965604697582e-06, + "logits/chosen": -2.3568668365478516, + "logits/rejected": -2.349731206893921, + "logps/chosen": -21.91687774658203, + "logps/rejected": -9.241042137145996, + "loss": 0.7078, + "rewards/accuracies": 0.78125, + "rewards/chosen": 0.14988428354263306, + "rewards/margins": 0.9810056686401367, + "rewards/rejected": -0.8311214447021484, + "step": 448 + }, + { + "epoch": 0.9857299670691547, + "grad_norm": 6.33512020111084, + "learning_rate": 3.0387321725463003e-06, + "logits/chosen": -2.307635545730591, + "logits/rejected": -2.2977371215820312, + "logps/chosen": -17.04680633544922, + "logps/rejected": -9.844083786010742, + "loss": 0.7663, + "rewards/accuracies": 0.8125, + "rewards/chosen": 0.12432314455509186, + "rewards/margins": 0.5705389976501465, + "rewards/rejected": -0.4462158679962158, + "step": 449 + }, + { + "epoch": 0.9879253567508233, + "grad_norm": 5.216354846954346, + "learning_rate": 3.0292596805735275e-06, + "logits/chosen": -2.2773430347442627, + "logits/rejected": -2.349567413330078, + "logps/chosen": -18.42083740234375, + "logps/rejected": -8.746360778808594, + "loss": 0.7209, + "rewards/accuracies": 0.8125, + "rewards/chosen": 0.09063999354839325, + "rewards/margins": 0.7439864873886108, + "rewards/rejected": -0.6533465385437012, + "step": 450 + }, + { + "epoch": 0.9901207464324918, + "grad_norm": 5.72788143157959, + "learning_rate": 3.019779227044398e-06, + "logits/chosen": -2.26011061668396, + "logits/rejected": -2.315460443496704, + "logps/chosen": -19.606307983398438, + "logps/rejected": -12.85045051574707, + "loss": 0.8855, + "rewards/accuracies": 0.8125, + "rewards/chosen": 0.05291711166501045, + "rewards/margins": 0.5900472402572632, + "rewards/rejected": -0.5371301770210266, + "step": 451 + }, + { + "epoch": 0.9923161361141603, + "grad_norm": 4.74468469619751, + "learning_rate": 3.0102909545716395e-06, + "logits/chosen": -2.370033025741577, + "logits/rejected": -2.3583390712738037, + "logps/chosen": -19.241247177124023, + "logps/rejected": -13.944446563720703, + "loss": 0.878, + "rewards/accuracies": 0.71875, + "rewards/chosen": 0.06830352544784546, + "rewards/margins": 0.6422482132911682, + "rewards/rejected": -0.5739446878433228, + "step": 452 + }, + { + "epoch": 0.9945115257958288, + "grad_norm": 4.331366539001465, + "learning_rate": 3.000795005885594e-06, + "logits/chosen": -2.29002046585083, + "logits/rejected": -2.368769884109497, + "logps/chosen": -19.165603637695312, + "logps/rejected": -14.322245597839355, + "loss": 0.846, + "rewards/accuracies": 0.8125, + "rewards/chosen": 0.0362529531121254, + "rewards/margins": 0.645837664604187, + "rewards/rejected": -0.6095846891403198, + "step": 453 + }, + { + "epoch": 0.9967069154774972, + "grad_norm": 6.133034706115723, + "learning_rate": 2.9912915238320755e-06, + "logits/chosen": -2.343209743499756, + "logits/rejected": -2.2830896377563477, + "logps/chosen": -20.217222213745117, + "logps/rejected": -15.0652437210083, + "loss": 0.9206, + "rewards/accuracies": 0.78125, + "rewards/chosen": 0.018072601407766342, + "rewards/margins": 0.5864449739456177, + "rewards/rejected": -0.568372368812561, + "step": 454 + }, + { + "epoch": 0.9989023051591658, + "grad_norm": 5.871572971343994, + "learning_rate": 2.9817806513702247e-06, + "logits/chosen": -2.2338972091674805, + "logits/rejected": -2.330247402191162, + "logps/chosen": -24.56080436706543, + "logps/rejected": -14.950778007507324, + "loss": 0.8771, + "rewards/accuracies": 0.875, + "rewards/chosen": 0.06272906064987183, + "rewards/margins": 0.8400812149047852, + "rewards/rejected": -0.7773522138595581, + "step": 455 + }, + { + "epoch": 1.0010976948408343, + "grad_norm": 6.872170448303223, + "learning_rate": 2.9722625315703514e-06, + "logits/chosen": -2.334846258163452, + "logits/rejected": -2.396892786026001, + "logps/chosen": -20.765228271484375, + "logps/rejected": -13.301582336425781, + "loss": 0.9065, + "rewards/accuracies": 0.75, + "rewards/chosen": 0.02812388725578785, + "rewards/margins": 0.5520917177200317, + "rewards/rejected": -0.5239678025245667, + "step": 456 + }, + { + "epoch": 1.0032930845225028, + "grad_norm": 5.5031962394714355, + "learning_rate": 2.9627373076117864e-06, + "logits/chosen": -2.308184862136841, + "logits/rejected": -2.3489980697631836, + "logps/chosen": -17.902585983276367, + "logps/rejected": -10.119697570800781, + "loss": 0.7064, + "rewards/accuracies": 0.90625, + "rewards/chosen": 0.1337275505065918, + "rewards/margins": 0.9021998643875122, + "rewards/rejected": -0.7684723138809204, + "step": 457 + }, + { + "epoch": 1.0054884742041712, + "grad_norm": 5.081422805786133, + "learning_rate": 2.9532051227807294e-06, + "logits/chosen": -2.3426356315612793, + "logits/rejected": -2.446756362915039, + "logps/chosen": -24.83384132385254, + "logps/rejected": -15.476089477539062, + "loss": 0.9477, + "rewards/accuracies": 0.6875, + "rewards/chosen": -0.038049302995204926, + "rewards/margins": 0.6294887661933899, + "rewards/rejected": -0.6675381660461426, + "step": 458 + }, + { + "epoch": 1.0076838638858396, + "grad_norm": 7.958917140960693, + "learning_rate": 2.9436661204680884e-06, + "logits/chosen": -2.304771900177002, + "logits/rejected": -2.409733295440674, + "logps/chosen": -19.9411678314209, + "logps/rejected": -13.451245307922363, + "loss": 0.9428, + "rewards/accuracies": 0.8125, + "rewards/chosen": -0.07250344753265381, + "rewards/margins": 0.6922086477279663, + "rewards/rejected": -0.7647121548652649, + "step": 459 + }, + { + "epoch": 1.0098792535675083, + "grad_norm": 7.695209503173828, + "learning_rate": 2.9341204441673267e-06, + "logits/chosen": -2.4130144119262695, + "logits/rejected": -2.3211066722869873, + "logps/chosen": -21.808847427368164, + "logps/rejected": -10.129591941833496, + "loss": 0.6938, + "rewards/accuracies": 0.875, + "rewards/chosen": 0.06284459680318832, + "rewards/margins": 1.0137550830841064, + "rewards/rejected": -0.9509105086326599, + "step": 460 + }, + { + "epoch": 1.0120746432491767, + "grad_norm": 8.558310508728027, + "learning_rate": 2.9245682374723015e-06, + "logits/chosen": -2.363746404647827, + "logits/rejected": -2.2999260425567627, + "logps/chosen": -23.764137268066406, + "logps/rejected": -12.646890640258789, + "loss": 0.8464, + "rewards/accuracies": 0.84375, + "rewards/chosen": -0.07450981438159943, + "rewards/margins": 0.649944543838501, + "rewards/rejected": -0.7244543433189392, + "step": 461 + }, + { + "epoch": 1.0142700329308452, + "grad_norm": 6.969803810119629, + "learning_rate": 2.915009644075111e-06, + "logits/chosen": -2.345712900161743, + "logits/rejected": -2.2984023094177246, + "logps/chosen": -23.32240867614746, + "logps/rejected": -12.624139785766602, + "loss": 0.7552, + "rewards/accuracies": 0.8125, + "rewards/chosen": 0.07847782224416733, + "rewards/margins": 0.8715475797653198, + "rewards/rejected": -0.7930698394775391, + "step": 462 + }, + { + "epoch": 1.0164654226125138, + "grad_norm": 6.514275074005127, + "learning_rate": 2.9054448077639193e-06, + "logits/chosen": -2.321444272994995, + "logits/rejected": -2.360914707183838, + "logps/chosen": -18.839618682861328, + "logps/rejected": -11.413566589355469, + "loss": 0.94, + "rewards/accuracies": 0.71875, + "rewards/chosen": 0.017170608043670654, + "rewards/margins": 0.463826984167099, + "rewards/rejected": -0.44665634632110596, + "step": 463 + }, + { + "epoch": 1.0186608122941823, + "grad_norm": 6.006606578826904, + "learning_rate": 2.8958738724208074e-06, + "logits/chosen": -2.3388967514038086, + "logits/rejected": -2.3111119270324707, + "logps/chosen": -23.030719757080078, + "logps/rejected": -13.477852821350098, + "loss": 0.7711, + "rewards/accuracies": 0.875, + "rewards/chosen": -0.03722686320543289, + "rewards/margins": 1.0160772800445557, + "rewards/rejected": -1.0533041954040527, + "step": 464 + }, + { + "epoch": 1.0208562019758507, + "grad_norm": 5.020242214202881, + "learning_rate": 2.8862969820196017e-06, + "logits/chosen": -2.3123841285705566, + "logits/rejected": -2.272430181503296, + "logps/chosen": -19.88169288635254, + "logps/rejected": -10.677448272705078, + "loss": 0.7777, + "rewards/accuracies": 0.8125, + "rewards/chosen": 0.03892090916633606, + "rewards/margins": 0.676476001739502, + "rewards/rejected": -0.6375551223754883, + "step": 465 + }, + { + "epoch": 1.0230515916575191, + "grad_norm": 14.649626731872559, + "learning_rate": 2.876714280623708e-06, + "logits/chosen": -2.3485302925109863, + "logits/rejected": -2.300239086151123, + "logps/chosen": -22.369789123535156, + "logps/rejected": -13.951744079589844, + "loss": 0.8978, + "rewards/accuracies": 0.6875, + "rewards/chosen": -0.06319882720708847, + "rewards/margins": 0.724540114402771, + "rewards/rejected": -0.7877388596534729, + "step": 466 + }, + { + "epoch": 1.0252469813391878, + "grad_norm": 12.121737480163574, + "learning_rate": 2.8671259123839473e-06, + "logits/chosen": -2.2911033630371094, + "logits/rejected": -2.3053789138793945, + "logps/chosen": -20.78561019897461, + "logps/rejected": -14.06844425201416, + "loss": 0.978, + "rewards/accuracies": 0.65625, + "rewards/chosen": -0.02713637985289097, + "rewards/margins": 0.5895265936851501, + "rewards/rejected": -0.6166629791259766, + "step": 467 + }, + { + "epoch": 1.0274423710208562, + "grad_norm": 5.312811851501465, + "learning_rate": 2.8575320215363855e-06, + "logits/chosen": -2.39201021194458, + "logits/rejected": -2.3016114234924316, + "logps/chosen": -17.093791961669922, + "logps/rejected": -11.01317024230957, + "loss": 0.8375, + "rewards/accuracies": 0.75, + "rewards/chosen": 0.012173615396022797, + "rewards/margins": 0.5083686113357544, + "rewards/rejected": -0.4961949586868286, + "step": 468 + }, + { + "epoch": 1.0296377607025247, + "grad_norm": 3.480792284011841, + "learning_rate": 2.847932752400164e-06, + "logits/chosen": -2.2379236221313477, + "logits/rejected": -2.272181987762451, + "logps/chosen": -26.447717666625977, + "logps/rejected": -10.113715171813965, + "loss": 0.6806, + "rewards/accuracies": 0.90625, + "rewards/chosen": 0.05238833278417587, + "rewards/margins": 1.4347848892211914, + "rewards/rejected": -1.3823965787887573, + "step": 469 + }, + { + "epoch": 1.031833150384193, + "grad_norm": 6.378970146179199, + "learning_rate": 2.8383282493753282e-06, + "logits/chosen": -2.3209898471832275, + "logits/rejected": -2.3701822757720947, + "logps/chosen": -25.746612548828125, + "logps/rejected": -13.422050476074219, + "loss": 0.8186, + "rewards/accuracies": 0.78125, + "rewards/chosen": 0.07319897413253784, + "rewards/margins": 0.9246252179145813, + "rewards/rejected": -0.8514262437820435, + "step": 470 + }, + { + "epoch": 1.0340285400658618, + "grad_norm": 9.738311767578125, + "learning_rate": 2.828718656940657e-06, + "logits/chosen": -2.292196035385132, + "logits/rejected": -2.2930097579956055, + "logps/chosen": -17.366518020629883, + "logps/rejected": -11.297025680541992, + "loss": 0.8188, + "rewards/accuracies": 0.6875, + "rewards/chosen": 0.05388388782739639, + "rewards/margins": 0.6057333946228027, + "rewards/rejected": -0.5518494844436646, + "step": 471 + }, + { + "epoch": 1.0362239297475302, + "grad_norm": 5.969666957855225, + "learning_rate": 2.8191041196514874e-06, + "logits/chosen": -2.2960739135742188, + "logits/rejected": -2.3564860820770264, + "logps/chosen": -22.130290985107422, + "logps/rejected": -16.745349884033203, + "loss": 0.8903, + "rewards/accuracies": 0.84375, + "rewards/chosen": 0.13813051581382751, + "rewards/margins": 0.9751704335212708, + "rewards/rejected": -0.8370399475097656, + "step": 472 + }, + { + "epoch": 1.0384193194291986, + "grad_norm": 8.502330780029297, + "learning_rate": 2.8094847821375417e-06, + "logits/chosen": -2.3107662200927734, + "logits/rejected": -2.247957944869995, + "logps/chosen": -20.09488296508789, + "logps/rejected": -13.177092552185059, + "loss": 0.8619, + "rewards/accuracies": 0.6875, + "rewards/chosen": -0.05494283139705658, + "rewards/margins": 0.6814887523651123, + "rewards/rejected": -0.7364315390586853, + "step": 473 + }, + { + "epoch": 1.0406147091108673, + "grad_norm": 4.86586332321167, + "learning_rate": 2.7998607891007495e-06, + "logits/chosen": -2.2360305786132812, + "logits/rejected": -2.351999044418335, + "logps/chosen": -25.469884872436523, + "logps/rejected": -12.238964080810547, + "loss": 0.7518, + "rewards/accuracies": 0.84375, + "rewards/chosen": 0.03077607974410057, + "rewards/margins": 1.022268295288086, + "rewards/rejected": -0.9914922714233398, + "step": 474 + }, + { + "epoch": 1.0428100987925357, + "grad_norm": 5.323648929595947, + "learning_rate": 2.7902322853130758e-06, + "logits/chosen": -2.3493967056274414, + "logits/rejected": -2.3672752380371094, + "logps/chosen": -21.324352264404297, + "logps/rejected": -12.99924087524414, + "loss": 0.7872, + "rewards/accuracies": 0.9375, + "rewards/chosen": 0.051297396421432495, + "rewards/margins": 0.9318706393241882, + "rewards/rejected": -0.8805732727050781, + "step": 475 + }, + { + "epoch": 1.0450054884742042, + "grad_norm": 9.397939682006836, + "learning_rate": 2.780599415614338e-06, + "logits/chosen": -2.343372344970703, + "logits/rejected": -2.3458945751190186, + "logps/chosen": -20.28595733642578, + "logps/rejected": -11.33232307434082, + "loss": 0.8225, + "rewards/accuracies": 0.8125, + "rewards/chosen": 0.057800546288490295, + "rewards/margins": 0.7880180478096008, + "rewards/rejected": -0.7302175164222717, + "step": 476 + }, + { + "epoch": 1.0472008781558726, + "grad_norm": 7.2292375564575195, + "learning_rate": 2.770962324910027e-06, + "logits/chosen": -2.3727457523345947, + "logits/rejected": -2.3275270462036133, + "logps/chosen": -23.036954879760742, + "logps/rejected": -11.280477523803711, + "loss": 0.7194, + "rewards/accuracies": 0.90625, + "rewards/chosen": 0.027677491307258606, + "rewards/margins": 0.8231653571128845, + "rewards/rejected": -0.7954878807067871, + "step": 477 + }, + { + "epoch": 1.0493962678375413, + "grad_norm": 8.571611404418945, + "learning_rate": 2.761321158169134e-06, + "logits/chosen": -2.2707719802856445, + "logits/rejected": -2.3974571228027344, + "logps/chosen": -19.268749237060547, + "logps/rejected": -10.586587905883789, + "loss": 0.8184, + "rewards/accuracies": 0.75, + "rewards/chosen": 0.004132624715566635, + "rewards/margins": 0.7264975309371948, + "rewards/rejected": -0.7223648428916931, + "step": 478 + }, + { + "epoch": 1.0515916575192097, + "grad_norm": 10.267123222351074, + "learning_rate": 2.7516760604219616e-06, + "logits/chosen": -2.3141329288482666, + "logits/rejected": -2.3562588691711426, + "logps/chosen": -23.546676635742188, + "logps/rejected": -10.247743606567383, + "loss": 0.7269, + "rewards/accuracies": 0.71875, + "rewards/chosen": -0.13976307213306427, + "rewards/margins": 1.0857172012329102, + "rewards/rejected": -1.225480318069458, + "step": 479 + }, + { + "epoch": 1.0537870472008781, + "grad_norm": 8.011184692382812, + "learning_rate": 2.742027176757948e-06, + "logits/chosen": -2.337991714477539, + "logits/rejected": -2.365306854248047, + "logps/chosen": -23.819690704345703, + "logps/rejected": -10.067850112915039, + "loss": 0.6582, + "rewards/accuracies": 0.84375, + "rewards/chosen": 0.08188200742006302, + "rewards/margins": 1.1291248798370361, + "rewards/rejected": -1.0472428798675537, + "step": 480 + }, + { + "epoch": 1.0559824368825466, + "grad_norm": 8.678403854370117, + "learning_rate": 2.7323746523234813e-06, + "logits/chosen": -2.260739803314209, + "logits/rejected": -2.315016269683838, + "logps/chosen": -28.939285278320312, + "logps/rejected": -15.770125389099121, + "loss": 0.9155, + "rewards/accuracies": 0.71875, + "rewards/chosen": 0.04919245094060898, + "rewards/margins": 0.9738469123840332, + "rewards/rejected": -0.924654483795166, + "step": 481 + }, + { + "epoch": 1.0581778265642152, + "grad_norm": 5.742441177368164, + "learning_rate": 2.7227186323197164e-06, + "logits/chosen": -2.3153116703033447, + "logits/rejected": -2.391371250152588, + "logps/chosen": -19.665122985839844, + "logps/rejected": -10.558384895324707, + "loss": 0.7369, + "rewards/accuracies": 0.75, + "rewards/chosen": 0.14078299701213837, + "rewards/margins": 0.8545687198638916, + "rewards/rejected": -0.7137857675552368, + "step": 482 + }, + { + "epoch": 1.0603732162458837, + "grad_norm": 6.094995975494385, + "learning_rate": 2.713059262000392e-06, + "logits/chosen": -2.381730318069458, + "logits/rejected": -2.3332326412200928, + "logps/chosen": -20.83930206298828, + "logps/rejected": -15.890161514282227, + "loss": 0.9044, + "rewards/accuracies": 0.78125, + "rewards/chosen": 0.09247031062841415, + "rewards/margins": 0.6750046610832214, + "rewards/rejected": -0.5825343728065491, + "step": 483 + }, + { + "epoch": 1.062568605927552, + "grad_norm": 5.33577823638916, + "learning_rate": 2.703396686669646e-06, + "logits/chosen": -2.2977828979492188, + "logits/rejected": -2.346433639526367, + "logps/chosen": -24.69493865966797, + "logps/rejected": -10.94466781616211, + "loss": 0.6964, + "rewards/accuracies": 0.875, + "rewards/chosen": -0.03435507044196129, + "rewards/margins": 1.0729467868804932, + "rewards/rejected": -1.107301950454712, + "step": 484 + }, + { + "epoch": 1.0647639956092205, + "grad_norm": 8.83690071105957, + "learning_rate": 2.6937310516798276e-06, + "logits/chosen": -2.3194973468780518, + "logits/rejected": -2.332828998565674, + "logps/chosen": -21.928625106811523, + "logps/rejected": -13.1103515625, + "loss": 0.8291, + "rewards/accuracies": 0.875, + "rewards/chosen": -0.10120926052331924, + "rewards/margins": 0.782268226146698, + "rewards/rejected": -0.8834775686264038, + "step": 485 + }, + { + "epoch": 1.0669593852908892, + "grad_norm": 9.110899925231934, + "learning_rate": 2.6840625024293123e-06, + "logits/chosen": -2.2905545234680176, + "logits/rejected": -2.280731201171875, + "logps/chosen": -20.927894592285156, + "logps/rejected": -15.54442024230957, + "loss": 0.849, + "rewards/accuracies": 0.84375, + "rewards/chosen": 0.034153182059526443, + "rewards/margins": 0.8824621438980103, + "rewards/rejected": -0.8483089804649353, + "step": 486 + }, + { + "epoch": 1.0691547749725576, + "grad_norm": 6.2910943031311035, + "learning_rate": 2.6743911843603134e-06, + "logits/chosen": -2.2246713638305664, + "logits/rejected": -2.320910930633545, + "logps/chosen": -24.896484375, + "logps/rejected": -7.879002571105957, + "loss": 0.7118, + "rewards/accuracies": 0.84375, + "rewards/chosen": 0.11002375185489655, + "rewards/margins": 0.7733902931213379, + "rewards/rejected": -0.6633665561676025, + "step": 487 + }, + { + "epoch": 1.071350164654226, + "grad_norm": 10.10474967956543, + "learning_rate": 2.6647172429566963e-06, + "logits/chosen": -2.281175136566162, + "logits/rejected": -2.279855728149414, + "logps/chosen": -18.865497589111328, + "logps/rejected": -14.626907348632812, + "loss": 0.9312, + "rewards/accuracies": 0.78125, + "rewards/chosen": -0.19256648421287537, + "rewards/margins": 0.4999960660934448, + "rewards/rejected": -0.6925625205039978, + "step": 488 + }, + { + "epoch": 1.0735455543358947, + "grad_norm": 7.026412487030029, + "learning_rate": 2.6550408237417886e-06, + "logits/chosen": -2.346653938293457, + "logits/rejected": -2.329099178314209, + "logps/chosen": -24.167951583862305, + "logps/rejected": -11.686748504638672, + "loss": 0.7787, + "rewards/accuracies": 0.75, + "rewards/chosen": -0.015237412415444851, + "rewards/margins": 0.867202639579773, + "rewards/rejected": -0.8824400901794434, + "step": 489 + }, + { + "epoch": 1.0757409440175631, + "grad_norm": 5.65709114074707, + "learning_rate": 2.6453620722761897e-06, + "logits/chosen": -2.3161253929138184, + "logits/rejected": -2.260692834854126, + "logps/chosen": -22.770465850830078, + "logps/rejected": -14.608541488647461, + "loss": 0.866, + "rewards/accuracies": 0.78125, + "rewards/chosen": -0.07928504794836044, + "rewards/margins": 0.6451736688613892, + "rewards/rejected": -0.7244588136672974, + "step": 490 + }, + { + "epoch": 1.0779363336992316, + "grad_norm": 7.390451431274414, + "learning_rate": 2.635681134155585e-06, + "logits/chosen": -2.3062405586242676, + "logits/rejected": -2.351980447769165, + "logps/chosen": -24.109310150146484, + "logps/rejected": -14.89892292022705, + "loss": 0.8688, + "rewards/accuracies": 0.75, + "rewards/chosen": 0.0690557062625885, + "rewards/margins": 1.0943150520324707, + "rewards/rejected": -1.0252593755722046, + "step": 491 + }, + { + "epoch": 1.0801317233809, + "grad_norm": 9.753083229064941, + "learning_rate": 2.6259981550085505e-06, + "logits/chosen": -2.3199896812438965, + "logits/rejected": -2.30045747756958, + "logps/chosen": -23.072227478027344, + "logps/rejected": -11.335121154785156, + "loss": 0.8415, + "rewards/accuracies": 0.6875, + "rewards/chosen": -0.10801962018013, + "rewards/margins": 0.9226139187812805, + "rewards/rejected": -1.0306334495544434, + "step": 492 + }, + { + "epoch": 1.0823271130625687, + "grad_norm": 9.559391021728516, + "learning_rate": 2.6163132804943675e-06, + "logits/chosen": -2.361140489578247, + "logits/rejected": -2.3496603965759277, + "logps/chosen": -20.47613525390625, + "logps/rejected": -11.543803215026855, + "loss": 0.7832, + "rewards/accuracies": 0.78125, + "rewards/chosen": -0.050104692578315735, + "rewards/margins": 0.7556217908859253, + "rewards/rejected": -0.8057264685630798, + "step": 493 + }, + { + "epoch": 1.0845225027442371, + "grad_norm": 6.683259963989258, + "learning_rate": 2.606626656300827e-06, + "logits/chosen": -2.323789596557617, + "logits/rejected": -2.271223545074463, + "logps/chosen": -20.67462158203125, + "logps/rejected": -11.763118743896484, + "loss": 0.7572, + "rewards/accuracies": 0.875, + "rewards/chosen": 0.07084425538778305, + "rewards/margins": 0.8640992045402527, + "rewards/rejected": -0.7932549715042114, + "step": 494 + }, + { + "epoch": 1.0867178924259056, + "grad_norm": 6.216057300567627, + "learning_rate": 2.5969384281420425e-06, + "logits/chosen": -2.255970001220703, + "logits/rejected": -2.2368826866149902, + "logps/chosen": -15.501781463623047, + "logps/rejected": -10.052569389343262, + "loss": 0.8267, + "rewards/accuracies": 0.71875, + "rewards/chosen": 0.017311323434114456, + "rewards/margins": 0.4922070801258087, + "rewards/rejected": -0.47489574551582336, + "step": 495 + }, + { + "epoch": 1.088913282107574, + "grad_norm": 5.462358474731445, + "learning_rate": 2.587248741756253e-06, + "logits/chosen": -2.2871689796447754, + "logits/rejected": -2.261716842651367, + "logps/chosen": -27.40030860900879, + "logps/rejected": -14.105131149291992, + "loss": 0.7714, + "rewards/accuracies": 0.8125, + "rewards/chosen": -0.02072109468281269, + "rewards/margins": 0.8808303475379944, + "rewards/rejected": -0.901551365852356, + "step": 496 + }, + { + "epoch": 1.0911086717892426, + "grad_norm": 5.209628105163574, + "learning_rate": 2.5775577429036346e-06, + "logits/chosen": -2.295696496963501, + "logits/rejected": -2.422459602355957, + "logps/chosen": -23.314756393432617, + "logps/rejected": -9.764581680297852, + "loss": 0.6545, + "rewards/accuracies": 0.8125, + "rewards/chosen": 0.21466970443725586, + "rewards/margins": 1.2359107732772827, + "rewards/rejected": -1.0212409496307373, + "step": 497 + }, + { + "epoch": 1.093304061470911, + "grad_norm": 9.333959579467773, + "learning_rate": 2.5678655773641077e-06, + "logits/chosen": -2.302668571472168, + "logits/rejected": -2.286771535873413, + "logps/chosen": -21.88730239868164, + "logps/rejected": -12.551828384399414, + "loss": 0.7906, + "rewards/accuracies": 0.84375, + "rewards/chosen": -0.03755124658346176, + "rewards/margins": 0.8626418113708496, + "rewards/rejected": -0.9001929759979248, + "step": 498 + }, + { + "epoch": 1.0954994511525795, + "grad_norm": 6.7280988693237305, + "learning_rate": 2.558172390935141e-06, + "logits/chosen": -2.2580811977386475, + "logits/rejected": -2.3480772972106934, + "logps/chosen": -25.80786895751953, + "logps/rejected": -10.555074691772461, + "loss": 0.703, + "rewards/accuracies": 0.8125, + "rewards/chosen": 0.0884111076593399, + "rewards/margins": 1.0041056871414185, + "rewards/rejected": -0.9156945943832397, + "step": 499 + }, + { + "epoch": 1.0976948408342482, + "grad_norm": 6.179262638092041, + "learning_rate": 2.548478329429561e-06, + "logits/chosen": -2.338765859603882, + "logits/rejected": -2.2902755737304688, + "logps/chosen": -21.667165756225586, + "logps/rejected": -18.325084686279297, + "loss": 1.1209, + "rewards/accuracies": 0.625, + "rewards/chosen": -0.05905440077185631, + "rewards/margins": 0.27884072065353394, + "rewards/rejected": -0.33789512515068054, + "step": 500 + }, + { + "epoch": 1.0998902305159166, + "grad_norm": 4.881867408752441, + "learning_rate": 2.5387835386733585e-06, + "logits/chosen": -2.3326308727264404, + "logits/rejected": -2.3200454711914062, + "logps/chosen": -21.303903579711914, + "logps/rejected": -10.037727355957031, + "loss": 0.7292, + "rewards/accuracies": 0.78125, + "rewards/chosen": 0.054762884974479675, + "rewards/margins": 0.8834924697875977, + "rewards/rejected": -0.8287296295166016, + "step": 501 + }, + { + "epoch": 1.102085620197585, + "grad_norm": 6.90359354019165, + "learning_rate": 2.5290881645034932e-06, + "logits/chosen": -2.3469488620758057, + "logits/rejected": -2.380147933959961, + "logps/chosen": -23.74352264404297, + "logps/rejected": -12.889476776123047, + "loss": 0.796, + "rewards/accuracies": 0.78125, + "rewards/chosen": 0.0006383843719959259, + "rewards/margins": 1.1013463735580444, + "rewards/rejected": -1.1007080078125, + "step": 502 + }, + { + "epoch": 1.1042810098792535, + "grad_norm": 10.043181419372559, + "learning_rate": 2.519392352765701e-06, + "logits/chosen": -2.371044158935547, + "logits/rejected": -2.2376604080200195, + "logps/chosen": -19.703800201416016, + "logps/rejected": -14.118764877319336, + "loss": 0.8728, + "rewards/accuracies": 0.75, + "rewards/chosen": -0.07204356044530869, + "rewards/margins": 0.5561725497245789, + "rewards/rejected": -0.6282161474227905, + "step": 503 + }, + { + "epoch": 1.1064763995609221, + "grad_norm": 5.477917671203613, + "learning_rate": 2.509696249312301e-06, + "logits/chosen": -2.3331849575042725, + "logits/rejected": -2.27480411529541, + "logps/chosen": -22.197994232177734, + "logps/rejected": -10.162934303283691, + "loss": 0.7273, + "rewards/accuracies": 0.78125, + "rewards/chosen": 0.03636087104678154, + "rewards/margins": 0.9884530901908875, + "rewards/rejected": -0.9520922899246216, + "step": 504 + }, + { + "epoch": 1.1086717892425906, + "grad_norm": 8.44482135772705, + "learning_rate": 2.5e-06, + "logits/chosen": -2.308072566986084, + "logits/rejected": -2.3319051265716553, + "logps/chosen": -19.7982177734375, + "logps/rejected": -10.436334609985352, + "loss": 0.7722, + "rewards/accuracies": 0.78125, + "rewards/chosen": -0.0411079004406929, + "rewards/margins": 0.8647053837776184, + "rewards/rejected": -0.9058132171630859, + "step": 505 + }, + { + "epoch": 1.110867178924259, + "grad_norm": 9.539579391479492, + "learning_rate": 2.4903037506876996e-06, + "logits/chosen": -2.328544855117798, + "logits/rejected": -2.339698076248169, + "logps/chosen": -25.973371505737305, + "logps/rejected": -18.507919311523438, + "loss": 0.9432, + "rewards/accuracies": 0.8125, + "rewards/chosen": -0.030126571655273438, + "rewards/margins": 0.8929418325424194, + "rewards/rejected": -0.9230684041976929, + "step": 506 + }, + { + "epoch": 1.1130625686059274, + "grad_norm": 7.940929889678955, + "learning_rate": 2.4806076472343e-06, + "logits/chosen": -2.3939225673675537, + "logits/rejected": -2.307854652404785, + "logps/chosen": -24.44528579711914, + "logps/rejected": -16.028257369995117, + "loss": 0.8673, + "rewards/accuracies": 0.78125, + "rewards/chosen": 0.08280456066131592, + "rewards/margins": 0.9818367958068848, + "rewards/rejected": -0.8990321755409241, + "step": 507 + }, + { + "epoch": 1.115257958287596, + "grad_norm": 6.884702682495117, + "learning_rate": 2.470911835496508e-06, + "logits/chosen": -2.325016736984253, + "logits/rejected": -2.40616774559021, + "logps/chosen": -23.83535385131836, + "logps/rejected": -11.415084838867188, + "loss": 0.8301, + "rewards/accuracies": 0.78125, + "rewards/chosen": 0.01772160828113556, + "rewards/margins": 0.7353428602218628, + "rewards/rejected": -0.7176212668418884, + "step": 508 + }, + { + "epoch": 1.1174533479692645, + "grad_norm": 6.673171043395996, + "learning_rate": 2.461216461326642e-06, + "logits/chosen": -2.3892860412597656, + "logits/rejected": -2.316361427307129, + "logps/chosen": -24.04204559326172, + "logps/rejected": -11.071301460266113, + "loss": 0.6525, + "rewards/accuracies": 0.96875, + "rewards/chosen": 0.04661805182695389, + "rewards/margins": 1.2658956050872803, + "rewards/rejected": -1.2192775011062622, + "step": 509 + }, + { + "epoch": 1.119648737650933, + "grad_norm": 6.269791126251221, + "learning_rate": 2.4515216705704396e-06, + "logits/chosen": -2.35085129737854, + "logits/rejected": -2.3756144046783447, + "logps/chosen": -21.623994827270508, + "logps/rejected": -14.480419158935547, + "loss": 0.8887, + "rewards/accuracies": 0.71875, + "rewards/chosen": -0.05676385760307312, + "rewards/margins": 0.8314831256866455, + "rewards/rejected": -0.888247013092041, + "step": 510 + }, + { + "epoch": 1.1218441273326016, + "grad_norm": 6.86275577545166, + "learning_rate": 2.4418276090648596e-06, + "logits/chosen": -2.2903640270233154, + "logits/rejected": -2.328355073928833, + "logps/chosen": -22.3763484954834, + "logps/rejected": -11.86290168762207, + "loss": 0.7381, + "rewards/accuracies": 0.84375, + "rewards/chosen": 0.013300411403179169, + "rewards/margins": 1.0081902742385864, + "rewards/rejected": -0.9948898553848267, + "step": 511 + }, + { + "epoch": 1.12403951701427, + "grad_norm": 9.084625244140625, + "learning_rate": 2.432134422635893e-06, + "logits/chosen": -2.3267626762390137, + "logits/rejected": -2.2859044075012207, + "logps/chosen": -21.043107986450195, + "logps/rejected": -12.442046165466309, + "loss": 0.8559, + "rewards/accuracies": 0.78125, + "rewards/chosen": 0.019764907658100128, + "rewards/margins": 0.7053239345550537, + "rewards/rejected": -0.6855589747428894, + "step": 512 + }, + { + "epoch": 1.1262349066959385, + "grad_norm": 7.669501304626465, + "learning_rate": 2.422442257096366e-06, + "logits/chosen": -2.353550434112549, + "logits/rejected": -2.3722033500671387, + "logps/chosen": -22.001750946044922, + "logps/rejected": -9.754518508911133, + "loss": 0.7088, + "rewards/accuracies": 0.75, + "rewards/chosen": 0.12981222569942474, + "rewards/margins": 0.8668014407157898, + "rewards/rejected": -0.7369892597198486, + "step": 513 + }, + { + "epoch": 1.128430296377607, + "grad_norm": 10.363723754882812, + "learning_rate": 2.4127512582437486e-06, + "logits/chosen": -2.3143248558044434, + "logits/rejected": -2.3078842163085938, + "logps/chosen": -24.376005172729492, + "logps/rejected": -15.226461410522461, + "loss": 0.9326, + "rewards/accuracies": 0.75, + "rewards/chosen": -0.22472436726093292, + "rewards/margins": 0.6675853133201599, + "rewards/rejected": -0.8923096656799316, + "step": 514 + }, + { + "epoch": 1.1306256860592756, + "grad_norm": 7.7791266441345215, + "learning_rate": 2.403061571857958e-06, + "logits/chosen": -2.376467704772949, + "logits/rejected": -2.265462875366211, + "logps/chosen": -17.012428283691406, + "logps/rejected": -13.313913345336914, + "loss": 0.8913, + "rewards/accuracies": 0.6875, + "rewards/chosen": -0.014788463711738586, + "rewards/margins": 0.5971038341522217, + "rewards/rejected": -0.6118922829627991, + "step": 515 + }, + { + "epoch": 1.132821075740944, + "grad_norm": 5.934682846069336, + "learning_rate": 2.3933733436991734e-06, + "logits/chosen": -2.340254545211792, + "logits/rejected": -2.319061279296875, + "logps/chosen": -20.40915870666504, + "logps/rejected": -11.210052490234375, + "loss": 0.7675, + "rewards/accuracies": 0.75, + "rewards/chosen": 0.00013789162039756775, + "rewards/margins": 0.8642052412033081, + "rewards/rejected": -0.8640673756599426, + "step": 516 + }, + { + "epoch": 1.1350164654226125, + "grad_norm": 9.762900352478027, + "learning_rate": 2.3836867195056333e-06, + "logits/chosen": -2.3846359252929688, + "logits/rejected": -2.316649913787842, + "logps/chosen": -24.181659698486328, + "logps/rejected": -15.667963027954102, + "loss": 0.8784, + "rewards/accuracies": 0.78125, + "rewards/chosen": -0.052952706813812256, + "rewards/margins": 0.8329753875732422, + "rewards/rejected": -0.8859280347824097, + "step": 517 + }, + { + "epoch": 1.137211855104281, + "grad_norm": 10.596639633178711, + "learning_rate": 2.3740018449914503e-06, + "logits/chosen": -2.3260669708251953, + "logits/rejected": -2.3258838653564453, + "logps/chosen": -14.412958145141602, + "logps/rejected": -12.238800048828125, + "loss": 0.9653, + "rewards/accuracies": 0.6875, + "rewards/chosen": -0.06726135313510895, + "rewards/margins": 0.36023619771003723, + "rewards/rejected": -0.4274975657463074, + "step": 518 + }, + { + "epoch": 1.1394072447859496, + "grad_norm": 10.4237060546875, + "learning_rate": 2.364318865844416e-06, + "logits/chosen": -2.347930908203125, + "logits/rejected": -2.337660789489746, + "logps/chosen": -20.771535873413086, + "logps/rejected": -12.997071266174316, + "loss": 0.8524, + "rewards/accuracies": 0.71875, + "rewards/chosen": -0.01133589819073677, + "rewards/margins": 0.700576901435852, + "rewards/rejected": -0.7119128108024597, + "step": 519 + }, + { + "epoch": 1.141602634467618, + "grad_norm": 6.661622524261475, + "learning_rate": 2.3546379277238107e-06, + "logits/chosen": -2.3014602661132812, + "logits/rejected": -2.375547170639038, + "logps/chosen": -21.250028610229492, + "logps/rejected": -10.202247619628906, + "loss": 0.731, + "rewards/accuracies": 0.84375, + "rewards/chosen": 0.04235329478979111, + "rewards/margins": 0.9067080020904541, + "rewards/rejected": -0.8643547296524048, + "step": 520 + }, + { + "epoch": 1.1437980241492864, + "grad_norm": 8.288076400756836, + "learning_rate": 2.344959176258212e-06, + "logits/chosen": -2.34618878364563, + "logits/rejected": -2.2824316024780273, + "logps/chosen": -16.397390365600586, + "logps/rejected": -10.269210815429688, + "loss": 0.8453, + "rewards/accuracies": 0.71875, + "rewards/chosen": 0.02188560925424099, + "rewards/margins": 0.44358524680137634, + "rewards/rejected": -0.4216996133327484, + "step": 521 + }, + { + "epoch": 1.145993413830955, + "grad_norm": 6.047231197357178, + "learning_rate": 2.3352827570433036e-06, + "logits/chosen": -2.299895763397217, + "logits/rejected": -2.3093342781066895, + "logps/chosen": -30.019575119018555, + "logps/rejected": -11.355772018432617, + "loss": 0.7292, + "rewards/accuracies": 0.8125, + "rewards/chosen": 0.07779978215694427, + "rewards/margins": 0.951260507106781, + "rewards/rejected": -0.8734606504440308, + "step": 522 + }, + { + "epoch": 1.1481888035126235, + "grad_norm": 5.04451322555542, + "learning_rate": 2.325608815639687e-06, + "logits/chosen": -2.3001108169555664, + "logits/rejected": -2.3524928092956543, + "logps/chosen": -19.051753997802734, + "logps/rejected": -10.565093994140625, + "loss": 0.8371, + "rewards/accuracies": 0.5625, + "rewards/chosen": -0.022858357056975365, + "rewards/margins": 0.646570086479187, + "rewards/rejected": -0.6694284677505493, + "step": 523 + }, + { + "epoch": 1.150384193194292, + "grad_norm": 6.81455135345459, + "learning_rate": 2.315937497570688e-06, + "logits/chosen": -2.2939376831054688, + "logits/rejected": -2.2317380905151367, + "logps/chosen": -19.80864906311035, + "logps/rejected": -11.135587692260742, + "loss": 0.8128, + "rewards/accuracies": 0.71875, + "rewards/chosen": 0.011634349822998047, + "rewards/margins": 0.7349785566329956, + "rewards/rejected": -0.7233442068099976, + "step": 524 + }, + { + "epoch": 1.1525795828759604, + "grad_norm": 8.786965370178223, + "learning_rate": 2.3062689483201732e-06, + "logits/chosen": -2.340775489807129, + "logits/rejected": -2.346869945526123, + "logps/chosen": -19.153776168823242, + "logps/rejected": -10.415740966796875, + "loss": 0.8269, + "rewards/accuracies": 0.75, + "rewards/chosen": 0.0716131180524826, + "rewards/margins": 0.688749372959137, + "rewards/rejected": -0.6171362996101379, + "step": 525 + }, + { + "epoch": 1.154774972557629, + "grad_norm": 7.3247504234313965, + "learning_rate": 2.296603313330355e-06, + "logits/chosen": -2.322134017944336, + "logits/rejected": -2.403510808944702, + "logps/chosen": -22.072208404541016, + "logps/rejected": -10.395179748535156, + "loss": 0.7544, + "rewards/accuracies": 0.8125, + "rewards/chosen": 0.06496301293373108, + "rewards/margins": 0.9994459748268127, + "rewards/rejected": -0.9344829320907593, + "step": 526 + }, + { + "epoch": 1.1569703622392975, + "grad_norm": 7.874275207519531, + "learning_rate": 2.286940737999609e-06, + "logits/chosen": -2.3437623977661133, + "logits/rejected": -2.403930902481079, + "logps/chosen": -25.250911712646484, + "logps/rejected": -15.07078742980957, + "loss": 0.9019, + "rewards/accuracies": 0.75, + "rewards/chosen": 0.02051650732755661, + "rewards/margins": 0.8506998419761658, + "rewards/rejected": -0.8301832675933838, + "step": 527 + }, + { + "epoch": 1.159165751920966, + "grad_norm": 6.287486553192139, + "learning_rate": 2.2772813676802853e-06, + "logits/chosen": -2.2481837272644043, + "logits/rejected": -2.338935136795044, + "logps/chosen": -21.099597930908203, + "logps/rejected": -12.632410049438477, + "loss": 0.7874, + "rewards/accuracies": 0.78125, + "rewards/chosen": 0.09598539769649506, + "rewards/margins": 0.7895666360855103, + "rewards/rejected": -0.6935812830924988, + "step": 528 + }, + { + "epoch": 1.1613611416026344, + "grad_norm": 8.335880279541016, + "learning_rate": 2.26762534767652e-06, + "logits/chosen": -2.4691050052642822, + "logits/rejected": -2.3216023445129395, + "logps/chosen": -17.575027465820312, + "logps/rejected": -10.989261627197266, + "loss": 0.7922, + "rewards/accuracies": 0.6875, + "rewards/chosen": 0.04628712683916092, + "rewards/margins": 0.693164587020874, + "rewards/rejected": -0.6468774080276489, + "step": 529 + }, + { + "epoch": 1.163556531284303, + "grad_norm": 4.961236000061035, + "learning_rate": 2.2579728232420524e-06, + "logits/chosen": -2.2612202167510986, + "logits/rejected": -2.2529544830322266, + "logps/chosen": -15.751286506652832, + "logps/rejected": -9.382078170776367, + "loss": 0.7733, + "rewards/accuracies": 0.84375, + "rewards/chosen": -0.0075488463044166565, + "rewards/margins": 0.70013028383255, + "rewards/rejected": -0.7076791524887085, + "step": 530 + }, + { + "epoch": 1.1657519209659715, + "grad_norm": 7.967287063598633, + "learning_rate": 2.248323939578039e-06, + "logits/chosen": -2.268648147583008, + "logits/rejected": -2.280182123184204, + "logps/chosen": -20.400197982788086, + "logps/rejected": -11.145841598510742, + "loss": 0.7626, + "rewards/accuracies": 0.75, + "rewards/chosen": 0.0876256674528122, + "rewards/margins": 0.9056125283241272, + "rewards/rejected": -0.8179869055747986, + "step": 531 + }, + { + "epoch": 1.16794731064764, + "grad_norm": 6.0232720375061035, + "learning_rate": 2.238678841830867e-06, + "logits/chosen": -2.2084717750549316, + "logits/rejected": -2.261118173599243, + "logps/chosen": -18.923809051513672, + "logps/rejected": -10.254813194274902, + "loss": 0.8045, + "rewards/accuracies": 0.6875, + "rewards/chosen": 0.0969836637377739, + "rewards/margins": 0.6044824123382568, + "rewards/rejected": -0.5074987411499023, + "step": 532 + }, + { + "epoch": 1.1701427003293086, + "grad_norm": 6.991965293884277, + "learning_rate": 2.229037675089974e-06, + "logits/chosen": -2.3134584426879883, + "logits/rejected": -2.276124954223633, + "logps/chosen": -18.349306106567383, + "logps/rejected": -9.643857955932617, + "loss": 0.7149, + "rewards/accuracies": 0.875, + "rewards/chosen": 0.11668010801076889, + "rewards/margins": 0.9437568187713623, + "rewards/rejected": -0.8270766735076904, + "step": 533 + }, + { + "epoch": 1.172338090010977, + "grad_norm": 9.253371238708496, + "learning_rate": 2.2194005843856637e-06, + "logits/chosen": -2.2831485271453857, + "logits/rejected": -2.3371338844299316, + "logps/chosen": -23.073291778564453, + "logps/rejected": -14.077463150024414, + "loss": 0.8691, + "rewards/accuracies": 0.71875, + "rewards/chosen": -0.019108561798930168, + "rewards/margins": 0.7781455516815186, + "rewards/rejected": -0.7972540855407715, + "step": 534 + }, + { + "epoch": 1.1745334796926454, + "grad_norm": 8.683618545532227, + "learning_rate": 2.2097677146869242e-06, + "logits/chosen": -2.362706184387207, + "logits/rejected": -2.3297712802886963, + "logps/chosen": -22.538930892944336, + "logps/rejected": -14.297994613647461, + "loss": 0.8558, + "rewards/accuracies": 0.71875, + "rewards/chosen": -0.02579372748732567, + "rewards/margins": 0.5780220627784729, + "rewards/rejected": -0.6038157939910889, + "step": 535 + }, + { + "epoch": 1.1767288693743139, + "grad_norm": 13.473367691040039, + "learning_rate": 2.2001392108992505e-06, + "logits/chosen": -2.261984348297119, + "logits/rejected": -2.257131338119507, + "logps/chosen": -18.18301773071289, + "logps/rejected": -13.25670337677002, + "loss": 0.9159, + "rewards/accuracies": 0.78125, + "rewards/chosen": -0.04601767659187317, + "rewards/margins": 0.5289838314056396, + "rewards/rejected": -0.57500159740448, + "step": 536 + }, + { + "epoch": 1.1789242590559825, + "grad_norm": 7.633896827697754, + "learning_rate": 2.1905152178624596e-06, + "logits/chosen": -2.2835702896118164, + "logits/rejected": -2.2883293628692627, + "logps/chosen": -19.89462661743164, + "logps/rejected": -12.626238822937012, + "loss": 0.7883, + "rewards/accuracies": 0.84375, + "rewards/chosen": 0.1262565553188324, + "rewards/margins": 0.6851190328598022, + "rewards/rejected": -0.558862566947937, + "step": 537 + }, + { + "epoch": 1.181119648737651, + "grad_norm": 7.112298011779785, + "learning_rate": 2.1808958803485134e-06, + "logits/chosen": -2.311741828918457, + "logits/rejected": -2.2555291652679443, + "logps/chosen": -20.535423278808594, + "logps/rejected": -11.617327690124512, + "loss": 0.8249, + "rewards/accuracies": 0.78125, + "rewards/chosen": -0.017695598304271698, + "rewards/margins": 0.5967564582824707, + "rewards/rejected": -0.6144520044326782, + "step": 538 + }, + { + "epoch": 1.1833150384193194, + "grad_norm": 5.003750324249268, + "learning_rate": 2.1712813430593436e-06, + "logits/chosen": -2.3198728561401367, + "logits/rejected": -2.3427395820617676, + "logps/chosen": -25.28256607055664, + "logps/rejected": -9.883782386779785, + "loss": 0.7371, + "rewards/accuracies": 0.875, + "rewards/chosen": 0.09716814756393433, + "rewards/margins": 0.940900981426239, + "rewards/rejected": -0.8437328338623047, + "step": 539 + }, + { + "epoch": 1.1855104281009878, + "grad_norm": 6.4724016189575195, + "learning_rate": 2.161671750624673e-06, + "logits/chosen": -2.2420482635498047, + "logits/rejected": -2.3570146560668945, + "logps/chosen": -21.55780792236328, + "logps/rejected": -15.548563957214355, + "loss": 0.9049, + "rewards/accuracies": 0.78125, + "rewards/chosen": 0.07957489043474197, + "rewards/margins": 0.7103015184402466, + "rewards/rejected": -0.63072669506073, + "step": 540 + }, + { + "epoch": 1.1877058177826565, + "grad_norm": 5.605111122131348, + "learning_rate": 2.1520672475998374e-06, + "logits/chosen": -2.3919334411621094, + "logits/rejected": -2.2979228496551514, + "logps/chosen": -15.924934387207031, + "logps/rejected": -11.273942947387695, + "loss": 0.8035, + "rewards/accuracies": 0.75, + "rewards/chosen": 0.1292649507522583, + "rewards/margins": 0.5840333700180054, + "rewards/rejected": -0.4547683596611023, + "step": 541 + }, + { + "epoch": 1.189901207464325, + "grad_norm": 5.747645854949951, + "learning_rate": 2.1424679784636145e-06, + "logits/chosen": -2.2977070808410645, + "logits/rejected": -2.301708221435547, + "logps/chosen": -20.587650299072266, + "logps/rejected": -11.416675567626953, + "loss": 0.8022, + "rewards/accuracies": 0.8125, + "rewards/chosen": 0.04250403493642807, + "rewards/margins": 0.796831488609314, + "rewards/rejected": -0.7543275356292725, + "step": 542 + }, + { + "epoch": 1.1920965971459934, + "grad_norm": 4.512691497802734, + "learning_rate": 2.1328740876160526e-06, + "logits/chosen": -2.2530736923217773, + "logits/rejected": -2.3274102210998535, + "logps/chosen": -22.72617530822754, + "logps/rejected": -13.50900650024414, + "loss": 0.865, + "rewards/accuracies": 0.78125, + "rewards/chosen": 0.13451239466667175, + "rewards/margins": 0.7616719007492065, + "rewards/rejected": -0.6271594762802124, + "step": 543 + }, + { + "epoch": 1.194291986827662, + "grad_norm": 7.769556522369385, + "learning_rate": 2.1232857193762923e-06, + "logits/chosen": -2.2638072967529297, + "logits/rejected": -2.3178606033325195, + "logps/chosen": -20.884693145751953, + "logps/rejected": -13.3298921585083, + "loss": 0.8588, + "rewards/accuracies": 0.71875, + "rewards/chosen": 0.03638541325926781, + "rewards/margins": 0.7506407499313354, + "rewards/rejected": -0.7142553329467773, + "step": 544 + }, + { + "epoch": 1.1964873765093305, + "grad_norm": 4.210056781768799, + "learning_rate": 2.113703017980399e-06, + "logits/chosen": -2.268528938293457, + "logits/rejected": -2.2505099773406982, + "logps/chosen": -21.2979736328125, + "logps/rejected": -10.145201683044434, + "loss": 0.806, + "rewards/accuracies": 0.78125, + "rewards/chosen": 0.13092830777168274, + "rewards/margins": 0.9798430800437927, + "rewards/rejected": -0.8489147424697876, + "step": 545 + }, + { + "epoch": 1.198682766190999, + "grad_norm": 5.240859508514404, + "learning_rate": 2.1041261275791934e-06, + "logits/chosen": -2.2415339946746826, + "logits/rejected": -2.353085517883301, + "logps/chosen": -20.664676666259766, + "logps/rejected": -10.272144317626953, + "loss": 0.7605, + "rewards/accuracies": 0.71875, + "rewards/chosen": 0.052175022661685944, + "rewards/margins": 0.8831397294998169, + "rewards/rejected": -0.8309646844863892, + "step": 546 + }, + { + "epoch": 1.2008781558726673, + "grad_norm": 6.940246105194092, + "learning_rate": 2.094555192236082e-06, + "logits/chosen": -2.352752685546875, + "logits/rejected": -2.2998435497283936, + "logps/chosen": -26.30046844482422, + "logps/rejected": -10.936307907104492, + "loss": 0.7374, + "rewards/accuracies": 0.8125, + "rewards/chosen": 0.08240009099245071, + "rewards/margins": 0.9008955955505371, + "rewards/rejected": -0.818495512008667, + "step": 547 + }, + { + "epoch": 1.203073545554336, + "grad_norm": 10.477368354797363, + "learning_rate": 2.0849903559248907e-06, + "logits/chosen": -2.240131378173828, + "logits/rejected": -2.348433494567871, + "logps/chosen": -20.58401107788086, + "logps/rejected": -10.452065467834473, + "loss": 0.6818, + "rewards/accuracies": 0.84375, + "rewards/chosen": -0.016450654715299606, + "rewards/margins": 0.9601907134056091, + "rewards/rejected": -0.9766413569450378, + "step": 548 + }, + { + "epoch": 1.2052689352360044, + "grad_norm": 10.613191604614258, + "learning_rate": 2.0754317625276984e-06, + "logits/chosen": -2.3052978515625, + "logits/rejected": -2.2949910163879395, + "logps/chosen": -20.250463485717773, + "logps/rejected": -11.173296928405762, + "loss": 0.8378, + "rewards/accuracies": 0.71875, + "rewards/chosen": -0.03195466846227646, + "rewards/margins": 0.7589899301528931, + "rewards/rejected": -0.7909445762634277, + "step": 549 + }, + { + "epoch": 1.2074643249176729, + "grad_norm": 8.828267097473145, + "learning_rate": 2.0658795558326745e-06, + "logits/chosen": -2.3525280952453613, + "logits/rejected": -2.3149402141571045, + "logps/chosen": -22.400985717773438, + "logps/rejected": -10.98021125793457, + "loss": 0.8096, + "rewards/accuracies": 0.75, + "rewards/chosen": 0.013297423720359802, + "rewards/margins": 0.7317354083061218, + "rewards/rejected": -0.7184380292892456, + "step": 550 + }, + { + "epoch": 1.2096597145993413, + "grad_norm": 5.943295001983643, + "learning_rate": 2.0563338795319124e-06, + "logits/chosen": -2.3683629035949707, + "logits/rejected": -2.348921298980713, + "logps/chosen": -23.856887817382812, + "logps/rejected": -11.961200714111328, + "loss": 0.7774, + "rewards/accuracies": 0.875, + "rewards/chosen": 0.051212988793849945, + "rewards/margins": 1.0052385330200195, + "rewards/rejected": -0.9540256261825562, + "step": 551 + }, + { + "epoch": 1.21185510428101, + "grad_norm": 7.031575679779053, + "learning_rate": 2.0467948772192714e-06, + "logits/chosen": -2.3540563583374023, + "logits/rejected": -2.2964532375335693, + "logps/chosen": -17.445579528808594, + "logps/rejected": -10.830303192138672, + "loss": 0.8042, + "rewards/accuracies": 0.8125, + "rewards/chosen": 0.005727390758693218, + "rewards/margins": 0.5892314910888672, + "rewards/rejected": -0.5835040807723999, + "step": 552 + }, + { + "epoch": 1.2140504939626784, + "grad_norm": 9.141935348510742, + "learning_rate": 2.037262692388214e-06, + "logits/chosen": -2.287322998046875, + "logits/rejected": -2.34805965423584, + "logps/chosen": -21.445873260498047, + "logps/rejected": -12.735130310058594, + "loss": 0.8022, + "rewards/accuracies": 0.78125, + "rewards/chosen": 0.012165870517492294, + "rewards/margins": 0.8326160311698914, + "rewards/rejected": -0.8204501271247864, + "step": 553 + }, + { + "epoch": 1.2162458836443468, + "grad_norm": 6.443539142608643, + "learning_rate": 2.02773746842965e-06, + "logits/chosen": -2.239424705505371, + "logits/rejected": -2.341888904571533, + "logps/chosen": -20.87397575378418, + "logps/rejected": -12.136924743652344, + "loss": 0.8178, + "rewards/accuracies": 0.75, + "rewards/chosen": -0.05236458778381348, + "rewards/margins": 0.7391312718391418, + "rewards/rejected": -0.7914958596229553, + "step": 554 + }, + { + "epoch": 1.2184412733260155, + "grad_norm": 10.701168060302734, + "learning_rate": 2.0182193486297757e-06, + "logits/chosen": -2.269219160079956, + "logits/rejected": -2.2648439407348633, + "logps/chosen": -22.858243942260742, + "logps/rejected": -13.833395957946777, + "loss": 0.8898, + "rewards/accuracies": 0.6875, + "rewards/chosen": -0.20513306558132172, + "rewards/margins": 0.7103500366210938, + "rewards/rejected": -0.9154832363128662, + "step": 555 + }, + { + "epoch": 1.220636663007684, + "grad_norm": 8.362030982971191, + "learning_rate": 2.0087084761679245e-06, + "logits/chosen": -2.1924595832824707, + "logits/rejected": -2.3378076553344727, + "logps/chosen": -23.616836547851562, + "logps/rejected": -9.201102256774902, + "loss": 0.6844, + "rewards/accuracies": 0.84375, + "rewards/chosen": 0.013586243614554405, + "rewards/margins": 0.9738850593566895, + "rewards/rejected": -0.9602987766265869, + "step": 556 + }, + { + "epoch": 1.2228320526893524, + "grad_norm": 10.594383239746094, + "learning_rate": 1.9992049941144066e-06, + "logits/chosen": -2.3824400901794434, + "logits/rejected": -2.3197388648986816, + "logps/chosen": -25.58890151977539, + "logps/rejected": -15.33682632446289, + "loss": 0.8926, + "rewards/accuracies": 0.71875, + "rewards/chosen": -0.1638583093881607, + "rewards/margins": 0.9415387511253357, + "rewards/rejected": -1.1053969860076904, + "step": 557 + }, + { + "epoch": 1.2250274423710208, + "grad_norm": 6.956945896148682, + "learning_rate": 1.9897090454283614e-06, + "logits/chosen": -2.3101649284362793, + "logits/rejected": -2.307924270629883, + "logps/chosen": -23.639381408691406, + "logps/rejected": -11.508820533752441, + "loss": 0.7774, + "rewards/accuracies": 0.78125, + "rewards/chosen": 0.01539243757724762, + "rewards/margins": 0.8538904190063477, + "rewards/rejected": -0.8384979963302612, + "step": 558 + }, + { + "epoch": 1.2272228320526892, + "grad_norm": 7.682737827301025, + "learning_rate": 1.9802207729556023e-06, + "logits/chosen": -2.3226852416992188, + "logits/rejected": -2.3344953060150146, + "logps/chosen": -20.801807403564453, + "logps/rejected": -11.750479698181152, + "loss": 0.9387, + "rewards/accuracies": 0.6875, + "rewards/chosen": -0.06511960923671722, + "rewards/margins": 0.5040086507797241, + "rewards/rejected": -0.5691282749176025, + "step": 559 + }, + { + "epoch": 1.2294182217343579, + "grad_norm": 9.384903907775879, + "learning_rate": 1.970740319426474e-06, + "logits/chosen": -2.332040548324585, + "logits/rejected": -2.3064050674438477, + "logps/chosen": -28.669025421142578, + "logps/rejected": -13.920568466186523, + "loss": 0.7993, + "rewards/accuracies": 0.84375, + "rewards/chosen": -0.05853881314396858, + "rewards/margins": 1.1536080837249756, + "rewards/rejected": -1.2121468782424927, + "step": 560 + }, + { + "epoch": 1.2316136114160263, + "grad_norm": 10.340078353881836, + "learning_rate": 1.9612678274537005e-06, + "logits/chosen": -2.397711753845215, + "logits/rejected": -2.359071731567383, + "logps/chosen": -27.721656799316406, + "logps/rejected": -15.664802551269531, + "loss": 0.9439, + "rewards/accuracies": 0.78125, + "rewards/chosen": -0.05386602506041527, + "rewards/margins": 0.7208966016769409, + "rewards/rejected": -0.7747626304626465, + "step": 561 + }, + { + "epoch": 1.2338090010976948, + "grad_norm": 6.03995943069458, + "learning_rate": 1.9518034395302413e-06, + "logits/chosen": -2.3363399505615234, + "logits/rejected": -2.278351306915283, + "logps/chosen": -23.042282104492188, + "logps/rejected": -14.336282730102539, + "loss": 0.8197, + "rewards/accuracies": 0.90625, + "rewards/chosen": 0.09039350599050522, + "rewards/margins": 0.8563604950904846, + "rewards/rejected": -0.7659670114517212, + "step": 562 + }, + { + "epoch": 1.2360043907793634, + "grad_norm": 8.038592338562012, + "learning_rate": 1.942347298027152e-06, + "logits/chosen": -2.2602267265319824, + "logits/rejected": -2.3177194595336914, + "logps/chosen": -28.35436248779297, + "logps/rejected": -10.782648086547852, + "loss": 0.7502, + "rewards/accuracies": 0.84375, + "rewards/chosen": 0.08170454949140549, + "rewards/margins": 0.8313093185424805, + "rewards/rejected": -0.7496048212051392, + "step": 563 + }, + { + "epoch": 1.2381997804610319, + "grad_norm": 5.56099796295166, + "learning_rate": 1.932899545191433e-06, + "logits/chosen": -2.277745246887207, + "logits/rejected": -2.3187880516052246, + "logps/chosen": -20.783117294311523, + "logps/rejected": -11.72138786315918, + "loss": 0.71, + "rewards/accuracies": 0.875, + "rewards/chosen": 0.12727870047092438, + "rewards/margins": 0.8532087802886963, + "rewards/rejected": -0.7259299755096436, + "step": 564 + }, + { + "epoch": 1.2403951701427003, + "grad_norm": 8.25932788848877, + "learning_rate": 1.9234603231439e-06, + "logits/chosen": -2.271496295928955, + "logits/rejected": -2.2847213745117188, + "logps/chosen": -22.986650466918945, + "logps/rejected": -15.193363189697266, + "loss": 0.9193, + "rewards/accuracies": 0.75, + "rewards/chosen": 0.010033978149294853, + "rewards/margins": 0.940314531326294, + "rewards/rejected": -0.9302806258201599, + "step": 565 + }, + { + "epoch": 1.242590559824369, + "grad_norm": 4.484811305999756, + "learning_rate": 1.914029773877039e-06, + "logits/chosen": -2.294851779937744, + "logits/rejected": -2.3359665870666504, + "logps/chosen": -20.862701416015625, + "logps/rejected": -10.026266098022461, + "loss": 0.7008, + "rewards/accuracies": 0.75, + "rewards/chosen": 0.052687376737594604, + "rewards/margins": 1.0147449970245361, + "rewards/rejected": -0.9620575308799744, + "step": 566 + }, + { + "epoch": 1.2447859495060374, + "grad_norm": 6.304704666137695, + "learning_rate": 1.9046080392528734e-06, + "logits/chosen": -2.200948715209961, + "logits/rejected": -2.248547077178955, + "logps/chosen": -24.4631404876709, + "logps/rejected": -12.562747955322266, + "loss": 0.815, + "rewards/accuracies": 0.78125, + "rewards/chosen": 0.06394200026988983, + "rewards/margins": 1.0856231451034546, + "rewards/rejected": -1.0216811895370483, + "step": 567 + }, + { + "epoch": 1.2469813391877058, + "grad_norm": 5.815040111541748, + "learning_rate": 1.895195261000831e-06, + "logits/chosen": -2.315217971801758, + "logits/rejected": -2.283458948135376, + "logps/chosen": -27.662736892700195, + "logps/rejected": -11.802858352661133, + "loss": 0.778, + "rewards/accuracies": 0.84375, + "rewards/chosen": -0.03663066029548645, + "rewards/margins": 0.9020322561264038, + "rewards/rejected": -0.9386628270149231, + "step": 568 + }, + { + "epoch": 1.2491767288693743, + "grad_norm": 12.249011993408203, + "learning_rate": 1.8857915807156092e-06, + "logits/chosen": -2.29628324508667, + "logits/rejected": -2.2696895599365234, + "logps/chosen": -24.545085906982422, + "logps/rejected": -12.295074462890625, + "loss": 0.8018, + "rewards/accuracies": 0.84375, + "rewards/chosen": 0.05770976096391678, + "rewards/margins": 0.8373286724090576, + "rewards/rejected": -0.7796189785003662, + "step": 569 + }, + { + "epoch": 1.2513721185510427, + "grad_norm": 11.68165397644043, + "learning_rate": 1.876397139855047e-06, + "logits/chosen": -2.3017168045043945, + "logits/rejected": -2.3384058475494385, + "logps/chosen": -19.50583267211914, + "logps/rejected": -12.806120872497559, + "loss": 0.9502, + "rewards/accuracies": 0.65625, + "rewards/chosen": -0.12223954498767853, + "rewards/margins": 0.6290686726570129, + "rewards/rejected": -0.7513082027435303, + "step": 570 + }, + { + "epoch": 1.2535675082327113, + "grad_norm": 5.761139869689941, + "learning_rate": 1.8670120797379958e-06, + "logits/chosen": -2.4278979301452637, + "logits/rejected": -2.332573890686035, + "logps/chosen": -21.713411331176758, + "logps/rejected": -10.450056076049805, + "loss": 0.7096, + "rewards/accuracies": 0.78125, + "rewards/chosen": 0.11197870224714279, + "rewards/margins": 0.790289580821991, + "rewards/rejected": -0.6783108711242676, + "step": 571 + }, + { + "epoch": 1.2557628979143798, + "grad_norm": 7.349134922027588, + "learning_rate": 1.8576365415421952e-06, + "logits/chosen": -2.2855048179626465, + "logits/rejected": -2.307605266571045, + "logps/chosen": -17.187366485595703, + "logps/rejected": -10.574321746826172, + "loss": 0.8193, + "rewards/accuracies": 0.78125, + "rewards/chosen": 0.07899583131074905, + "rewards/margins": 0.5769809484481812, + "rewards/rejected": -0.4979851543903351, + "step": 572 + }, + { + "epoch": 1.2579582875960482, + "grad_norm": 8.864078521728516, + "learning_rate": 1.8482706663021467e-06, + "logits/chosen": -2.316099166870117, + "logits/rejected": -2.4364302158355713, + "logps/chosen": -16.598718643188477, + "logps/rejected": -9.304157257080078, + "loss": 0.8036, + "rewards/accuracies": 0.8125, + "rewards/chosen": -0.023954614996910095, + "rewards/margins": 0.6026211977005005, + "rewards/rejected": -0.6265758275985718, + "step": 573 + }, + { + "epoch": 1.2601536772777169, + "grad_norm": 7.2920308113098145, + "learning_rate": 1.8389145949069953e-06, + "logits/chosen": -2.311638832092285, + "logits/rejected": -2.3084805011749268, + "logps/chosen": -24.171340942382812, + "logps/rejected": -11.01889419555664, + "loss": 0.7538, + "rewards/accuracies": 0.6875, + "rewards/chosen": -0.0005981642752885818, + "rewards/margins": 0.9836317300796509, + "rewards/rejected": -0.9842299818992615, + "step": 574 + }, + { + "epoch": 1.2623490669593853, + "grad_norm": 5.73350191116333, + "learning_rate": 1.8295684680984064e-06, + "logits/chosen": -2.3550052642822266, + "logits/rejected": -2.3727738857269287, + "logps/chosen": -19.880443572998047, + "logps/rejected": -11.300536155700684, + "loss": 0.7184, + "rewards/accuracies": 0.84375, + "rewards/chosen": 0.11995154619216919, + "rewards/margins": 0.9147524237632751, + "rewards/rejected": -0.7948008179664612, + "step": 575 + }, + { + "epoch": 1.2645444566410537, + "grad_norm": 7.762485027313232, + "learning_rate": 1.8202324264684546e-06, + "logits/chosen": -2.321082592010498, + "logits/rejected": -2.3335936069488525, + "logps/chosen": -22.020591735839844, + "logps/rejected": -11.06096363067627, + "loss": 0.7465, + "rewards/accuracies": 0.78125, + "rewards/chosen": 0.16837848722934723, + "rewards/margins": 0.8557047247886658, + "rewards/rejected": -0.6873262524604797, + "step": 576 + }, + { + "epoch": 1.2667398463227224, + "grad_norm": 7.1133880615234375, + "learning_rate": 1.8109066104575023e-06, + "logits/chosen": -2.25290846824646, + "logits/rejected": -2.291252374649048, + "logps/chosen": -16.12826919555664, + "logps/rejected": -13.576467514038086, + "loss": 0.9407, + "rewards/accuracies": 0.6875, + "rewards/chosen": -0.07207485288381577, + "rewards/margins": 0.5390644073486328, + "rewards/rejected": -0.6111392974853516, + "step": 577 + }, + { + "epoch": 1.2689352360043908, + "grad_norm": 6.930362224578857, + "learning_rate": 1.8015911603520895e-06, + "logits/chosen": -2.3748748302459717, + "logits/rejected": -2.2770352363586426, + "logps/chosen": -15.177204132080078, + "logps/rejected": -10.055876731872559, + "loss": 0.8065, + "rewards/accuracies": 0.8125, + "rewards/chosen": 0.10187117755413055, + "rewards/margins": 0.49988412857055664, + "rewards/rejected": -0.3980129361152649, + "step": 578 + }, + { + "epoch": 1.2711306256860593, + "grad_norm": 7.314796447753906, + "learning_rate": 1.7922862162828242e-06, + "logits/chosen": -2.4613027572631836, + "logits/rejected": -2.3669557571411133, + "logps/chosen": -21.86223602294922, + "logps/rejected": -11.79267692565918, + "loss": 0.8098, + "rewards/accuracies": 0.8125, + "rewards/chosen": 0.0410163439810276, + "rewards/margins": 0.8419078588485718, + "rewards/rejected": -0.8008914589881897, + "step": 579 + }, + { + "epoch": 1.2733260153677277, + "grad_norm": 7.840477466583252, + "learning_rate": 1.7829919182222752e-06, + "logits/chosen": -2.2705116271972656, + "logits/rejected": -2.3549280166625977, + "logps/chosen": -16.7214298248291, + "logps/rejected": -9.668839454650879, + "loss": 0.8443, + "rewards/accuracies": 0.78125, + "rewards/chosen": 0.006537839770317078, + "rewards/margins": 0.5482538342475891, + "rewards/rejected": -0.5417160391807556, + "step": 580 + }, + { + "epoch": 1.2755214050493962, + "grad_norm": 9.975746154785156, + "learning_rate": 1.7737084059828637e-06, + "logits/chosen": -2.308288097381592, + "logits/rejected": -2.2945263385772705, + "logps/chosen": -19.2215633392334, + "logps/rejected": -9.784078598022461, + "loss": 0.6994, + "rewards/accuracies": 0.84375, + "rewards/chosen": 0.10235311090946198, + "rewards/margins": 0.8730628490447998, + "rewards/rejected": -0.7707098126411438, + "step": 581 + }, + { + "epoch": 1.2777167947310648, + "grad_norm": 9.812601089477539, + "learning_rate": 1.764435819214762e-06, + "logits/chosen": -2.2570254802703857, + "logits/rejected": -2.370835304260254, + "logps/chosen": -24.492950439453125, + "logps/rejected": -13.760238647460938, + "loss": 0.8129, + "rewards/accuracies": 0.90625, + "rewards/chosen": 0.06407330185174942, + "rewards/margins": 0.9926495552062988, + "rewards/rejected": -0.9285762310028076, + "step": 582 + }, + { + "epoch": 1.2799121844127332, + "grad_norm": 6.761222839355469, + "learning_rate": 1.755174297403795e-06, + "logits/chosen": -2.2884902954101562, + "logits/rejected": -2.340023994445801, + "logps/chosen": -22.65121078491211, + "logps/rejected": -10.267145156860352, + "loss": 0.7036, + "rewards/accuracies": 0.78125, + "rewards/chosen": 0.020283987745642662, + "rewards/margins": 0.970897912979126, + "rewards/rejected": -0.9506139159202576, + "step": 583 + }, + { + "epoch": 1.2821075740944017, + "grad_norm": 5.213293075561523, + "learning_rate": 1.7459239798693365e-06, + "logits/chosen": -2.258117437362671, + "logits/rejected": -2.327763080596924, + "logps/chosen": -18.63202667236328, + "logps/rejected": -7.666975498199463, + "loss": 0.7205, + "rewards/accuracies": 0.75, + "rewards/chosen": 0.09631773084402084, + "rewards/margins": 0.7262113094329834, + "rewards/rejected": -0.6298935413360596, + "step": 584 + }, + { + "epoch": 1.2843029637760703, + "grad_norm": 6.437871932983398, + "learning_rate": 1.7366850057622176e-06, + "logits/chosen": -2.241556406021118, + "logits/rejected": -2.266603946685791, + "logps/chosen": -20.111963272094727, + "logps/rejected": -16.927959442138672, + "loss": 1.0025, + "rewards/accuracies": 0.78125, + "rewards/chosen": 0.09613069146871567, + "rewards/margins": 0.6213204860687256, + "rewards/rejected": -0.5251898765563965, + "step": 585 + }, + { + "epoch": 1.2864983534577388, + "grad_norm": 8.618307113647461, + "learning_rate": 1.7274575140626318e-06, + "logits/chosen": -2.2256720066070557, + "logits/rejected": -2.2476024627685547, + "logps/chosen": -21.752702713012695, + "logps/rejected": -8.02682113647461, + "loss": 0.6548, + "rewards/accuracies": 0.75, + "rewards/chosen": 0.11019796133041382, + "rewards/margins": 1.2175419330596924, + "rewards/rejected": -1.1073439121246338, + "step": 586 + }, + { + "epoch": 1.2886937431394072, + "grad_norm": 6.519025802612305, + "learning_rate": 1.7182416435780454e-06, + "logits/chosen": -2.305727243423462, + "logits/rejected": -2.298759698867798, + "logps/chosen": -22.402408599853516, + "logps/rejected": -9.142102241516113, + "loss": 0.6494, + "rewards/accuracies": 0.875, + "rewards/chosen": 0.13550861179828644, + "rewards/margins": 1.193606972694397, + "rewards/rejected": -1.058098316192627, + "step": 587 + }, + { + "epoch": 1.2908891328210759, + "grad_norm": 6.377670764923096, + "learning_rate": 1.7090375329411085e-06, + "logits/chosen": -2.2581775188446045, + "logits/rejected": -2.2949557304382324, + "logps/chosen": -23.584163665771484, + "logps/rejected": -10.748200416564941, + "loss": 0.7717, + "rewards/accuracies": 0.75, + "rewards/chosen": 0.08184422552585602, + "rewards/margins": 0.9663362503051758, + "rewards/rejected": -0.884492039680481, + "step": 588 + }, + { + "epoch": 1.2930845225027443, + "grad_norm": 7.334181308746338, + "learning_rate": 1.699845320607571e-06, + "logits/chosen": -2.3284082412719727, + "logits/rejected": -2.2626307010650635, + "logps/chosen": -23.804515838623047, + "logps/rejected": -14.898904800415039, + "loss": 0.8598, + "rewards/accuracies": 0.75, + "rewards/chosen": 0.05425605922937393, + "rewards/margins": 0.9250165224075317, + "rewards/rejected": -0.8707605004310608, + "step": 589 + }, + { + "epoch": 1.2952799121844127, + "grad_norm": 5.097620010375977, + "learning_rate": 1.6906651448541977e-06, + "logits/chosen": -2.2644309997558594, + "logits/rejected": -2.3991267681121826, + "logps/chosen": -21.256227493286133, + "logps/rejected": -10.035719871520996, + "loss": 0.7051, + "rewards/accuracies": 0.8125, + "rewards/chosen": 0.14272454380989075, + "rewards/margins": 1.0451469421386719, + "rewards/rejected": -0.9024224281311035, + "step": 590 + }, + { + "epoch": 1.2974753018660812, + "grad_norm": 8.508007049560547, + "learning_rate": 1.6814971437766892e-06, + "logits/chosen": -2.2637534141540527, + "logits/rejected": -2.227092742919922, + "logps/chosen": -18.255521774291992, + "logps/rejected": -10.249122619628906, + "loss": 0.8276, + "rewards/accuracies": 0.78125, + "rewards/chosen": -0.03183445334434509, + "rewards/margins": 0.5978387594223022, + "rewards/rejected": -0.629673182964325, + "step": 591 + }, + { + "epoch": 1.2996706915477496, + "grad_norm": 7.709940433502197, + "learning_rate": 1.6723414552876052e-06, + "logits/chosen": -2.308002471923828, + "logits/rejected": -2.250213623046875, + "logps/chosen": -20.619705200195312, + "logps/rejected": -14.285510063171387, + "loss": 0.8963, + "rewards/accuracies": 0.84375, + "rewards/chosen": 0.05097903311252594, + "rewards/margins": 0.5606150031089783, + "rewards/rejected": -0.5096359252929688, + "step": 592 + }, + { + "epoch": 1.3018660812294183, + "grad_norm": 6.3299946784973145, + "learning_rate": 1.6631982171142897e-06, + "logits/chosen": -2.261220693588257, + "logits/rejected": -2.283700942993164, + "logps/chosen": -23.349563598632812, + "logps/rejected": -10.86942195892334, + "loss": 0.7142, + "rewards/accuracies": 0.875, + "rewards/chosen": 0.08610594272613525, + "rewards/margins": 1.0204660892486572, + "rewards/rejected": -0.934360146522522, + "step": 593 + }, + { + "epoch": 1.3040614709110867, + "grad_norm": 7.599740028381348, + "learning_rate": 1.6540675667967976e-06, + "logits/chosen": -2.3450398445129395, + "logits/rejected": -2.2551536560058594, + "logps/chosen": -18.89197540283203, + "logps/rejected": -12.675716400146484, + "loss": 0.8504, + "rewards/accuracies": 0.65625, + "rewards/chosen": -0.08833372592926025, + "rewards/margins": 0.7258880138397217, + "rewards/rejected": -0.8142217397689819, + "step": 594 + }, + { + "epoch": 1.3062568605927551, + "grad_norm": 8.625227928161621, + "learning_rate": 1.6449496416858285e-06, + "logits/chosen": -2.2444748878479004, + "logits/rejected": -2.3313210010528564, + "logps/chosen": -24.318784713745117, + "logps/rejected": -12.065596580505371, + "loss": 0.7509, + "rewards/accuracies": 0.84375, + "rewards/chosen": 0.047348037362098694, + "rewards/margins": 1.075792908668518, + "rewards/rejected": -1.0284448862075806, + "step": 595 + }, + { + "epoch": 1.3084522502744238, + "grad_norm": 6.686420440673828, + "learning_rate": 1.6358445789406585e-06, + "logits/chosen": -2.3516736030578613, + "logits/rejected": -2.295524835586548, + "logps/chosen": -21.888641357421875, + "logps/rejected": -12.229887962341309, + "loss": 0.8185, + "rewards/accuracies": 0.78125, + "rewards/chosen": 0.05561601370573044, + "rewards/margins": 0.7821831703186035, + "rewards/rejected": -0.7265671491622925, + "step": 596 + }, + { + "epoch": 1.3106476399560922, + "grad_norm": 5.866522312164307, + "learning_rate": 1.6267525155270775e-06, + "logits/chosen": -2.321063995361328, + "logits/rejected": -2.345247268676758, + "logps/chosen": -22.20502471923828, + "logps/rejected": -10.635605812072754, + "loss": 0.7004, + "rewards/accuracies": 0.8125, + "rewards/chosen": 0.13774268329143524, + "rewards/margins": 1.0633516311645508, + "rewards/rejected": -0.9256089925765991, + "step": 597 + }, + { + "epoch": 1.3128430296377607, + "grad_norm": 11.989717483520508, + "learning_rate": 1.6176735882153284e-06, + "logits/chosen": -2.3791770935058594, + "logits/rejected": -2.339172124862671, + "logps/chosen": -20.8574161529541, + "logps/rejected": -12.953239440917969, + "loss": 0.83, + "rewards/accuracies": 0.8125, + "rewards/chosen": 0.0441550612449646, + "rewards/margins": 0.7832438945770264, + "rewards/rejected": -0.739088773727417, + "step": 598 + }, + { + "epoch": 1.3150384193194293, + "grad_norm": 9.708837509155273, + "learning_rate": 1.6086079335780502e-06, + "logits/chosen": -2.2837138175964355, + "logits/rejected": -2.2774574756622314, + "logps/chosen": -20.526235580444336, + "logps/rejected": -10.272727966308594, + "loss": 0.6816, + "rewards/accuracies": 0.78125, + "rewards/chosen": 0.0673438161611557, + "rewards/margins": 0.8778731822967529, + "rewards/rejected": -0.8105292320251465, + "step": 599 + }, + { + "epoch": 1.3172338090010978, + "grad_norm": 7.864263534545898, + "learning_rate": 1.5995556879882246e-06, + "logits/chosen": -2.356355667114258, + "logits/rejected": -2.3562049865722656, + "logps/chosen": -19.904991149902344, + "logps/rejected": -11.627508163452148, + "loss": 0.7446, + "rewards/accuracies": 0.875, + "rewards/chosen": 0.07776273787021637, + "rewards/margins": 0.837946891784668, + "rewards/rejected": -0.7601841688156128, + "step": 600 + }, + { + "epoch": 1.3194291986827662, + "grad_norm": 9.945621490478516, + "learning_rate": 1.5905169876171223e-06, + "logits/chosen": -2.2994587421417236, + "logits/rejected": -2.340704917907715, + "logps/chosen": -19.92719841003418, + "logps/rejected": -12.511524200439453, + "loss": 0.8377, + "rewards/accuracies": 0.71875, + "rewards/chosen": -0.06123330071568489, + "rewards/margins": 0.6635480523109436, + "rewards/rejected": -0.724781334400177, + "step": 601 + }, + { + "epoch": 1.3216245883644346, + "grad_norm": 11.670140266418457, + "learning_rate": 1.5814919684322544e-06, + "logits/chosen": -2.253532886505127, + "logits/rejected": -2.3085215091705322, + "logps/chosen": -22.0670166015625, + "logps/rejected": -15.236169815063477, + "loss": 0.9712, + "rewards/accuracies": 0.78125, + "rewards/chosen": -0.041367307305336, + "rewards/margins": 0.8814539313316345, + "rewards/rejected": -0.9228212237358093, + "step": 602 + }, + { + "epoch": 1.323819978046103, + "grad_norm": 6.727524757385254, + "learning_rate": 1.5724807661953329e-06, + "logits/chosen": -2.266571044921875, + "logits/rejected": -2.3808159828186035, + "logps/chosen": -21.894920349121094, + "logps/rejected": -9.660882949829102, + "loss": 0.7984, + "rewards/accuracies": 0.6875, + "rewards/chosen": -0.012565597891807556, + "rewards/margins": 0.8706260919570923, + "rewards/rejected": -0.8831916451454163, + "step": 603 + }, + { + "epoch": 1.3260153677277717, + "grad_norm": 11.664478302001953, + "learning_rate": 1.56348351646022e-06, + "logits/chosen": -2.3068418502807617, + "logits/rejected": -2.3269388675689697, + "logps/chosen": -22.625497817993164, + "logps/rejected": -11.628764152526855, + "loss": 0.8177, + "rewards/accuracies": 0.84375, + "rewards/chosen": 0.0710265040397644, + "rewards/margins": 0.6519577503204346, + "rewards/rejected": -0.5809313058853149, + "step": 604 + }, + { + "epoch": 1.3282107574094402, + "grad_norm": 8.326641082763672, + "learning_rate": 1.5545003545708942e-06, + "logits/chosen": -2.298421859741211, + "logits/rejected": -2.2954275608062744, + "logps/chosen": -22.19131851196289, + "logps/rejected": -10.66403865814209, + "loss": 0.7418, + "rewards/accuracies": 0.78125, + "rewards/chosen": 0.06426955759525299, + "rewards/margins": 1.0526033639907837, + "rewards/rejected": -0.9883337020874023, + "step": 605 + }, + { + "epoch": 1.3304061470911086, + "grad_norm": 6.182759761810303, + "learning_rate": 1.5455314156594126e-06, + "logits/chosen": -2.3587841987609863, + "logits/rejected": -2.344381332397461, + "logps/chosen": -21.562278747558594, + "logps/rejected": -11.864307403564453, + "loss": 0.7827, + "rewards/accuracies": 0.71875, + "rewards/chosen": -0.08919505774974823, + "rewards/margins": 0.8450158834457397, + "rewards/rejected": -0.9342109560966492, + "step": 606 + }, + { + "epoch": 1.3326015367727773, + "grad_norm": 10.691828727722168, + "learning_rate": 1.5365768346438798e-06, + "logits/chosen": -2.2707901000976562, + "logits/rejected": -2.3436715602874756, + "logps/chosen": -20.035856246948242, + "logps/rejected": -11.214956283569336, + "loss": 0.8423, + "rewards/accuracies": 0.8125, + "rewards/chosen": 0.01770446076989174, + "rewards/margins": 0.6999155282974243, + "rewards/rejected": -0.6822110414505005, + "step": 607 + }, + { + "epoch": 1.3347969264544457, + "grad_norm": 13.612374305725098, + "learning_rate": 1.5276367462264152e-06, + "logits/chosen": -2.1999752521514893, + "logits/rejected": -2.2819418907165527, + "logps/chosen": -16.572717666625977, + "logps/rejected": -8.16527271270752, + "loss": 0.7741, + "rewards/accuracies": 0.78125, + "rewards/chosen": 0.046401944011449814, + "rewards/margins": 0.754936933517456, + "rewards/rejected": -0.7085350155830383, + "step": 608 + }, + { + "epoch": 1.3369923161361141, + "grad_norm": 6.667941093444824, + "learning_rate": 1.5187112848911323e-06, + "logits/chosen": -2.3558480739593506, + "logits/rejected": -2.3914763927459717, + "logps/chosen": -20.603124618530273, + "logps/rejected": -9.974440574645996, + "loss": 0.7312, + "rewards/accuracies": 0.875, + "rewards/chosen": 0.09626440703868866, + "rewards/margins": 1.0095330476760864, + "rewards/rejected": -0.9132686257362366, + "step": 609 + }, + { + "epoch": 1.3391877058177828, + "grad_norm": 8.911206245422363, + "learning_rate": 1.509800584902108e-06, + "logits/chosen": -2.301635265350342, + "logits/rejected": -2.351564407348633, + "logps/chosen": -22.66827392578125, + "logps/rejected": -12.447919845581055, + "loss": 0.813, + "rewards/accuracies": 0.75, + "rewards/chosen": -0.027142884209752083, + "rewards/margins": 0.97141033411026, + "rewards/rejected": -0.9985532760620117, + "step": 610 + }, + { + "epoch": 1.3413830954994512, + "grad_norm": 5.421774864196777, + "learning_rate": 1.50090478030137e-06, + "logits/chosen": -2.2879533767700195, + "logits/rejected": -2.3026154041290283, + "logps/chosen": -21.532020568847656, + "logps/rejected": -10.294055938720703, + "loss": 0.7439, + "rewards/accuracies": 0.71875, + "rewards/chosen": 0.07867896556854248, + "rewards/margins": 0.9818447232246399, + "rewards/rejected": -0.9031658172607422, + "step": 611 + }, + { + "epoch": 1.3435784851811197, + "grad_norm": 6.579399108886719, + "learning_rate": 1.4920240049068749e-06, + "logits/chosen": -2.3022241592407227, + "logits/rejected": -2.291295051574707, + "logps/chosen": -21.16923713684082, + "logps/rejected": -9.061702728271484, + "loss": 0.7422, + "rewards/accuracies": 0.8125, + "rewards/chosen": 0.04894019663333893, + "rewards/margins": 0.8373510241508484, + "rewards/rejected": -0.7884107828140259, + "step": 612 + }, + { + "epoch": 1.345773874862788, + "grad_norm": 5.334139823913574, + "learning_rate": 1.4831583923105e-06, + "logits/chosen": -2.1957497596740723, + "logits/rejected": -2.2865476608276367, + "logps/chosen": -21.681808471679688, + "logps/rejected": -9.344619750976562, + "loss": 0.7113, + "rewards/accuracies": 0.8125, + "rewards/chosen": 0.17219513654708862, + "rewards/margins": 0.8527023792266846, + "rewards/rejected": -0.680507242679596, + "step": 613 + }, + { + "epoch": 1.3479692645444565, + "grad_norm": 11.850005149841309, + "learning_rate": 1.47430807587603e-06, + "logits/chosen": -2.2808213233947754, + "logits/rejected": -2.324070930480957, + "logps/chosen": -13.812728881835938, + "logps/rejected": -11.546747207641602, + "loss": 0.9288, + "rewards/accuracies": 0.59375, + "rewards/chosen": -0.06807052344083786, + "rewards/margins": 0.3613322675228119, + "rewards/rejected": -0.42940276861190796, + "step": 614 + }, + { + "epoch": 1.3501646542261252, + "grad_norm": 6.64136266708374, + "learning_rate": 1.4654731887371524e-06, + "logits/chosen": -2.2478113174438477, + "logits/rejected": -2.319643497467041, + "logps/chosen": -25.574260711669922, + "logps/rejected": -12.913862228393555, + "loss": 0.7369, + "rewards/accuracies": 0.90625, + "rewards/chosen": 0.1434711515903473, + "rewards/margins": 1.0417921543121338, + "rewards/rejected": -0.8983209133148193, + "step": 615 + }, + { + "epoch": 1.3523600439077936, + "grad_norm": 10.494634628295898, + "learning_rate": 1.4566538637954556e-06, + "logits/chosen": -2.214296340942383, + "logits/rejected": -2.2468421459198, + "logps/chosen": -19.69860076904297, + "logps/rejected": -10.450303077697754, + "loss": 0.7753, + "rewards/accuracies": 0.75, + "rewards/chosen": 0.0715034231543541, + "rewards/margins": 0.7445260286331177, + "rewards/rejected": -0.6730226278305054, + "step": 616 + }, + { + "epoch": 1.354555433589462, + "grad_norm": 7.629697799682617, + "learning_rate": 1.4478502337184274e-06, + "logits/chosen": -2.3394224643707275, + "logits/rejected": -2.2981069087982178, + "logps/chosen": -21.374021530151367, + "logps/rejected": -13.276947975158691, + "loss": 0.772, + "rewards/accuracies": 0.8125, + "rewards/chosen": 0.11859530210494995, + "rewards/margins": 0.8991226553916931, + "rewards/rejected": -0.7805273532867432, + "step": 617 + }, + { + "epoch": 1.3567508232711307, + "grad_norm": 6.685688018798828, + "learning_rate": 1.4390624309374618e-06, + "logits/chosen": -2.280592441558838, + "logits/rejected": -2.414875030517578, + "logps/chosen": -21.928415298461914, + "logps/rejected": -9.04220199584961, + "loss": 0.7474, + "rewards/accuracies": 0.78125, + "rewards/chosen": 0.09325066953897476, + "rewards/margins": 1.0403122901916504, + "rewards/rejected": -0.9470615386962891, + "step": 618 + }, + { + "epoch": 1.3589462129527992, + "grad_norm": 5.429327964782715, + "learning_rate": 1.430290587645865e-06, + "logits/chosen": -2.2931711673736572, + "logits/rejected": -2.3608384132385254, + "logps/chosen": -27.577693939208984, + "logps/rejected": -10.63676643371582, + "loss": 0.671, + "rewards/accuracies": 0.78125, + "rewards/chosen": 0.06562984734773636, + "rewards/margins": 1.1430988311767578, + "rewards/rejected": -1.077468991279602, + "step": 619 + }, + { + "epoch": 1.3611416026344676, + "grad_norm": 6.951731204986572, + "learning_rate": 1.421534835796867e-06, + "logits/chosen": -2.284036874771118, + "logits/rejected": -2.340190887451172, + "logps/chosen": -22.195070266723633, + "logps/rejected": -11.246692657470703, + "loss": 0.7731, + "rewards/accuracies": 0.78125, + "rewards/chosen": 0.07484187930822372, + "rewards/margins": 0.9202183485031128, + "rewards/rejected": -0.8453764915466309, + "step": 620 + }, + { + "epoch": 1.363336992316136, + "grad_norm": 5.5013604164123535, + "learning_rate": 1.4127953071016385e-06, + "logits/chosen": -2.3443572521209717, + "logits/rejected": -2.2324447631835938, + "logps/chosen": -24.410945892333984, + "logps/rejected": -10.45059585571289, + "loss": 0.6881, + "rewards/accuracies": 0.84375, + "rewards/chosen": 0.022320883348584175, + "rewards/margins": 1.1777617931365967, + "rewards/rejected": -1.1554409265518188, + "step": 621 + }, + { + "epoch": 1.3655323819978047, + "grad_norm": 6.671341419219971, + "learning_rate": 1.4040721330273063e-06, + "logits/chosen": -2.2495126724243164, + "logits/rejected": -2.3079583644866943, + "logps/chosen": -26.410785675048828, + "logps/rejected": -13.460811614990234, + "loss": 0.8601, + "rewards/accuracies": 0.71875, + "rewards/chosen": 0.017114847898483276, + "rewards/margins": 1.1208982467651367, + "rewards/rejected": -1.1037834882736206, + "step": 622 + }, + { + "epoch": 1.3677277716794731, + "grad_norm": 13.658313751220703, + "learning_rate": 1.395365444794981e-06, + "logits/chosen": -2.225109338760376, + "logits/rejected": -2.359435558319092, + "logps/chosen": -20.910913467407227, + "logps/rejected": -11.052978515625, + "loss": 0.7618, + "rewards/accuracies": 0.875, + "rewards/chosen": 0.18513336777687073, + "rewards/margins": 0.873185396194458, + "rewards/rejected": -0.6880519986152649, + "step": 623 + }, + { + "epoch": 1.3699231613611416, + "grad_norm": 6.782786846160889, + "learning_rate": 1.3866753733777766e-06, + "logits/chosen": -2.3307008743286133, + "logits/rejected": -2.3325958251953125, + "logps/chosen": -21.54385757446289, + "logps/rejected": -8.939606666564941, + "loss": 0.6691, + "rewards/accuracies": 0.71875, + "rewards/chosen": 0.09676853567361832, + "rewards/margins": 0.9515146017074585, + "rewards/rejected": -0.8547461032867432, + "step": 624 + }, + { + "epoch": 1.37211855104281, + "grad_norm": 6.9263410568237305, + "learning_rate": 1.3780020494988447e-06, + "logits/chosen": -2.266835927963257, + "logits/rejected": -2.290130376815796, + "logps/chosen": -25.256898880004883, + "logps/rejected": -9.844560623168945, + "loss": 0.7099, + "rewards/accuracies": 0.8125, + "rewards/chosen": 0.03770245611667633, + "rewards/margins": 1.209524393081665, + "rewards/rejected": -1.17182195186615, + "step": 625 + }, + { + "epoch": 1.3743139407244787, + "grad_norm": 5.386626243591309, + "learning_rate": 1.3693456036294061e-06, + "logits/chosen": -2.2897837162017822, + "logits/rejected": -2.296311855316162, + "logps/chosen": -16.78245735168457, + "logps/rejected": -9.544893264770508, + "loss": 0.7222, + "rewards/accuracies": 0.84375, + "rewards/chosen": 0.14815039932727814, + "rewards/margins": 0.7126171588897705, + "rewards/rejected": -0.5644667148590088, + "step": 626 + }, + { + "epoch": 1.376509330406147, + "grad_norm": 9.906569480895996, + "learning_rate": 1.3607061659867893e-06, + "logits/chosen": -2.2987747192382812, + "logits/rejected": -2.37379789352417, + "logps/chosen": -21.942153930664062, + "logps/rejected": -14.199853897094727, + "loss": 0.8431, + "rewards/accuracies": 0.75, + "rewards/chosen": 0.062028490006923676, + "rewards/margins": 0.9648635387420654, + "rewards/rejected": -0.9028350710868835, + "step": 627 + }, + { + "epoch": 1.3787047200878155, + "grad_norm": 6.001233100891113, + "learning_rate": 1.3520838665324704e-06, + "logits/chosen": -2.2744500637054443, + "logits/rejected": -2.299311637878418, + "logps/chosen": -19.875694274902344, + "logps/rejected": -9.716584205627441, + "loss": 0.7302, + "rewards/accuracies": 0.84375, + "rewards/chosen": 0.04750049114227295, + "rewards/margins": 0.8758196830749512, + "rewards/rejected": -0.8283191919326782, + "step": 628 + }, + { + "epoch": 1.3809001097694842, + "grad_norm": 7.546182632446289, + "learning_rate": 1.343478834970121e-06, + "logits/chosen": -2.3422470092773438, + "logits/rejected": -2.415135383605957, + "logps/chosen": -24.375473022460938, + "logps/rejected": -11.362316131591797, + "loss": 0.7589, + "rewards/accuracies": 0.75, + "rewards/chosen": 0.06067121401429176, + "rewards/margins": 0.989262580871582, + "rewards/rejected": -0.9285913109779358, + "step": 629 + }, + { + "epoch": 1.3830954994511526, + "grad_norm": 10.08180046081543, + "learning_rate": 1.3348912007436538e-06, + "logits/chosen": -2.2763381004333496, + "logits/rejected": -2.256406545639038, + "logps/chosen": -22.64297103881836, + "logps/rejected": -16.188823699951172, + "loss": 0.9282, + "rewards/accuracies": 0.8125, + "rewards/chosen": 0.1101061999797821, + "rewards/margins": 1.0070350170135498, + "rewards/rejected": -0.8969286680221558, + "step": 630 + }, + { + "epoch": 1.385290889132821, + "grad_norm": 12.735706329345703, + "learning_rate": 1.3263210930352737e-06, + "logits/chosen": -2.255270004272461, + "logits/rejected": -2.2508039474487305, + "logps/chosen": -24.61071014404297, + "logps/rejected": -14.970820426940918, + "loss": 0.8316, + "rewards/accuracies": 0.8125, + "rewards/chosen": 0.07555194199085236, + "rewards/margins": 1.0000357627868652, + "rewards/rejected": -0.9244839549064636, + "step": 631 + }, + { + "epoch": 1.3874862788144895, + "grad_norm": 7.744649887084961, + "learning_rate": 1.3177686407635418e-06, + "logits/chosen": -2.25099515914917, + "logits/rejected": -2.259213924407959, + "logps/chosen": -19.736896514892578, + "logps/rejected": -11.566102981567383, + "loss": 0.7553, + "rewards/accuracies": 0.75, + "rewards/chosen": 0.11749571561813354, + "rewards/margins": 0.8145424127578735, + "rewards/rejected": -0.6970466375350952, + "step": 632 + }, + { + "epoch": 1.3896816684961582, + "grad_norm": 6.42110013961792, + "learning_rate": 1.3092339725814293e-06, + "logits/chosen": -2.3431472778320312, + "logits/rejected": -2.3225836753845215, + "logps/chosen": -24.88370704650879, + "logps/rejected": -10.997604370117188, + "loss": 0.7178, + "rewards/accuracies": 0.78125, + "rewards/chosen": 0.056892167776823044, + "rewards/margins": 0.9540488123893738, + "rewards/rejected": -0.8971566557884216, + "step": 633 + }, + { + "epoch": 1.3918770581778266, + "grad_norm": 8.198354721069336, + "learning_rate": 1.3007172168743854e-06, + "logits/chosen": -2.2364721298217773, + "logits/rejected": -2.2169389724731445, + "logps/chosen": -29.71295738220215, + "logps/rejected": -17.012653350830078, + "loss": 0.8402, + "rewards/accuracies": 0.84375, + "rewards/chosen": -0.007015421986579895, + "rewards/margins": 1.1894762516021729, + "rewards/rejected": -1.1964917182922363, + "step": 634 + }, + { + "epoch": 1.394072447859495, + "grad_norm": 8.46764850616455, + "learning_rate": 1.2922185017584038e-06, + "logits/chosen": -2.264983654022217, + "logits/rejected": -2.271099090576172, + "logps/chosen": -21.81379508972168, + "logps/rejected": -11.220830917358398, + "loss": 0.7517, + "rewards/accuracies": 0.78125, + "rewards/chosen": 0.07613758742809296, + "rewards/margins": 0.9404584765434265, + "rewards/rejected": -0.8643208146095276, + "step": 635 + }, + { + "epoch": 1.3962678375411635, + "grad_norm": 6.605400085449219, + "learning_rate": 1.2837379550781005e-06, + "logits/chosen": -2.368940591812134, + "logits/rejected": -2.312143087387085, + "logps/chosen": -22.041481018066406, + "logps/rejected": -16.294395446777344, + "loss": 0.9895, + "rewards/accuracies": 0.75, + "rewards/chosen": 0.08214890956878662, + "rewards/margins": 0.5746477246284485, + "rewards/rejected": -0.4924987852573395, + "step": 636 + }, + { + "epoch": 1.3984632272228321, + "grad_norm": 8.978492736816406, + "learning_rate": 1.275275704404783e-06, + "logits/chosen": -2.2419798374176025, + "logits/rejected": -2.2099437713623047, + "logps/chosen": -27.026142120361328, + "logps/rejected": -12.915885925292969, + "loss": 0.7794, + "rewards/accuracies": 0.875, + "rewards/chosen": -0.00047025084495544434, + "rewards/margins": 1.3020012378692627, + "rewards/rejected": -1.30247163772583, + "step": 637 + }, + { + "epoch": 1.4006586169045006, + "grad_norm": 9.830564498901367, + "learning_rate": 1.266831877034537e-06, + "logits/chosen": -2.256542921066284, + "logits/rejected": -2.3181984424591064, + "logps/chosen": -26.209686279296875, + "logps/rejected": -11.520742416381836, + "loss": 0.6849, + "rewards/accuracies": 0.8125, + "rewards/chosen": 0.13590988516807556, + "rewards/margins": 1.2674816846847534, + "rewards/rejected": -1.1315717697143555, + "step": 638 + }, + { + "epoch": 1.402854006586169, + "grad_norm": 6.137882232666016, + "learning_rate": 1.2584065999863102e-06, + "logits/chosen": -2.284844398498535, + "logits/rejected": -2.3578481674194336, + "logps/chosen": -23.429718017578125, + "logps/rejected": -10.921028137207031, + "loss": 0.7569, + "rewards/accuracies": 0.8125, + "rewards/chosen": 0.07631979882717133, + "rewards/margins": 0.7318113446235657, + "rewards/rejected": -0.6554915308952332, + "step": 639 + }, + { + "epoch": 1.4050493962678376, + "grad_norm": 9.932879447937012, + "learning_rate": 1.2500000000000007e-06, + "logits/chosen": -2.3173117637634277, + "logits/rejected": -2.2574501037597656, + "logps/chosen": -22.321590423583984, + "logps/rejected": -14.684344291687012, + "loss": 0.8452, + "rewards/accuracies": 0.71875, + "rewards/chosen": -0.022527679800987244, + "rewards/margins": 0.91542649269104, + "rewards/rejected": -0.9379542469978333, + "step": 640 + }, + { + "epoch": 1.407244785949506, + "grad_norm": 11.732462882995605, + "learning_rate": 1.2416122035345506e-06, + "logits/chosen": -2.2233474254608154, + "logits/rejected": -2.350597858428955, + "logps/chosen": -21.614450454711914, + "logps/rejected": -13.806549072265625, + "loss": 0.8289, + "rewards/accuracies": 0.875, + "rewards/chosen": 0.02734570950269699, + "rewards/margins": 0.9277887344360352, + "rewards/rejected": -0.9004430770874023, + "step": 641 + }, + { + "epoch": 1.4094401756311745, + "grad_norm": 8.964897155761719, + "learning_rate": 1.2332433367660443e-06, + "logits/chosen": -2.3086395263671875, + "logits/rejected": -2.4242308139801025, + "logps/chosen": -20.400959014892578, + "logps/rejected": -15.17016887664795, + "loss": 0.9129, + "rewards/accuracies": 0.78125, + "rewards/chosen": -0.06498737633228302, + "rewards/margins": 0.6273730397224426, + "rewards/rejected": -0.6923604011535645, + "step": 642 + }, + { + "epoch": 1.411635565312843, + "grad_norm": 8.811767578125, + "learning_rate": 1.2248935255858118e-06, + "logits/chosen": -2.337164878845215, + "logits/rejected": -2.3390893936157227, + "logps/chosen": -20.301923751831055, + "logps/rejected": -13.365152359008789, + "loss": 0.874, + "rewards/accuracies": 0.71875, + "rewards/chosen": -0.01909365877509117, + "rewards/margins": 0.7496833801269531, + "rewards/rejected": -0.7687770128250122, + "step": 643 + }, + { + "epoch": 1.4138309549945114, + "grad_norm": 8.05325698852539, + "learning_rate": 1.2165628955985314e-06, + "logits/chosen": -2.37947154045105, + "logits/rejected": -2.363570213317871, + "logps/chosen": -23.159076690673828, + "logps/rejected": -11.53666877746582, + "loss": 0.74, + "rewards/accuracies": 0.84375, + "rewards/chosen": 0.05979155749082565, + "rewards/margins": 0.8970844745635986, + "rewards/rejected": -0.8372929096221924, + "step": 644 + }, + { + "epoch": 1.41602634467618, + "grad_norm": 11.32064437866211, + "learning_rate": 1.2082515721203429e-06, + "logits/chosen": -2.298936367034912, + "logits/rejected": -2.394646644592285, + "logps/chosen": -20.886850357055664, + "logps/rejected": -11.36214828491211, + "loss": 0.8877, + "rewards/accuracies": 0.71875, + "rewards/chosen": 0.035199642181396484, + "rewards/margins": 0.7127372026443481, + "rewards/rejected": -0.6775375604629517, + "step": 645 + }, + { + "epoch": 1.4182217343578485, + "grad_norm": 12.001448631286621, + "learning_rate": 1.1999596801769617e-06, + "logits/chosen": -2.354550838470459, + "logits/rejected": -2.300246238708496, + "logps/chosen": -20.45081901550293, + "logps/rejected": -16.67715072631836, + "loss": 1.0092, + "rewards/accuracies": 0.59375, + "rewards/chosen": -0.1403292864561081, + "rewards/margins": 0.6176349520683289, + "rewards/rejected": -0.7579641938209534, + "step": 646 + }, + { + "epoch": 1.420417124039517, + "grad_norm": 6.730834484100342, + "learning_rate": 1.1916873445017982e-06, + "logits/chosen": -2.408071517944336, + "logits/rejected": -2.3679609298706055, + "logps/chosen": -23.50677490234375, + "logps/rejected": -14.295957565307617, + "loss": 0.9044, + "rewards/accuracies": 0.78125, + "rewards/chosen": -0.035784341394901276, + "rewards/margins": 0.9361075162887573, + "rewards/rejected": -0.9718918800354004, + "step": 647 + }, + { + "epoch": 1.4226125137211856, + "grad_norm": 6.142631530761719, + "learning_rate": 1.1834346895340817e-06, + "logits/chosen": -2.283304214477539, + "logits/rejected": -2.324650287628174, + "logps/chosen": -19.739864349365234, + "logps/rejected": -11.370397567749023, + "loss": 0.773, + "rewards/accuracies": 0.75, + "rewards/chosen": 0.08815725892782211, + "rewards/margins": 0.864177942276001, + "rewards/rejected": -0.7760207653045654, + "step": 648 + }, + { + "epoch": 1.424807903402854, + "grad_norm": 7.696084499359131, + "learning_rate": 1.1752018394169882e-06, + "logits/chosen": -2.278122901916504, + "logits/rejected": -2.3596887588500977, + "logps/chosen": -21.272916793823242, + "logps/rejected": -11.957027435302734, + "loss": 0.8295, + "rewards/accuracies": 0.8125, + "rewards/chosen": -0.04064164683222771, + "rewards/margins": 0.7882517576217651, + "rewards/rejected": -0.8288934230804443, + "step": 649 + }, + { + "epoch": 1.4270032930845225, + "grad_norm": 8.252701759338379, + "learning_rate": 1.1669889179957725e-06, + "logits/chosen": -2.3437411785125732, + "logits/rejected": -2.293290138244629, + "logps/chosen": -18.37685775756836, + "logps/rejected": -12.420761108398438, + "loss": 0.8174, + "rewards/accuracies": 0.75, + "rewards/chosen": 0.01584087684750557, + "rewards/margins": 0.5727925300598145, + "rewards/rejected": -0.5569517016410828, + "step": 650 + }, + { + "epoch": 1.429198682766191, + "grad_norm": 8.367561340332031, + "learning_rate": 1.158796048815906e-06, + "logits/chosen": -2.334059238433838, + "logits/rejected": -2.3328163623809814, + "logps/chosen": -20.888351440429688, + "logps/rejected": -10.583667755126953, + "loss": 0.7693, + "rewards/accuracies": 0.8125, + "rewards/chosen": 0.01611892506480217, + "rewards/margins": 0.9742887616157532, + "rewards/rejected": -0.9581698179244995, + "step": 651 + }, + { + "epoch": 1.4313940724478595, + "grad_norm": 6.492325782775879, + "learning_rate": 1.1506233551212186e-06, + "logits/chosen": -2.2274742126464844, + "logits/rejected": -2.30332088470459, + "logps/chosen": -20.746814727783203, + "logps/rejected": -14.613504409790039, + "loss": 0.8862, + "rewards/accuracies": 0.8125, + "rewards/chosen": 0.05309009924530983, + "rewards/margins": 0.7812837362289429, + "rewards/rejected": -0.7281936407089233, + "step": 652 + }, + { + "epoch": 1.433589462129528, + "grad_norm": 7.120379447937012, + "learning_rate": 1.1424709598520426e-06, + "logits/chosen": -2.279757499694824, + "logits/rejected": -2.3859994411468506, + "logps/chosen": -18.22530746459961, + "logps/rejected": -8.492435455322266, + "loss": 0.7989, + "rewards/accuracies": 0.78125, + "rewards/chosen": 0.015699587762355804, + "rewards/margins": 0.8215601444244385, + "rewards/rejected": -0.8058606386184692, + "step": 653 + }, + { + "epoch": 1.4357848518111964, + "grad_norm": 12.999929428100586, + "learning_rate": 1.134338985643366e-06, + "logits/chosen": -2.2692627906799316, + "logits/rejected": -2.2525415420532227, + "logps/chosen": -20.028894424438477, + "logps/rejected": -9.751667022705078, + "loss": 0.711, + "rewards/accuracies": 0.84375, + "rewards/chosen": 0.1015341728925705, + "rewards/margins": 1.0113970041275024, + "rewards/rejected": -0.9098629355430603, + "step": 654 + }, + { + "epoch": 1.4379802414928649, + "grad_norm": 7.106282711029053, + "learning_rate": 1.1262275548229852e-06, + "logits/chosen": -2.313683032989502, + "logits/rejected": -2.3203182220458984, + "logps/chosen": -24.759023666381836, + "logps/rejected": -11.700056076049805, + "loss": 0.7551, + "rewards/accuracies": 0.75, + "rewards/chosen": 0.12376014143228531, + "rewards/margins": 1.090596318244934, + "rewards/rejected": -0.9668360352516174, + "step": 655 + }, + { + "epoch": 1.4401756311745335, + "grad_norm": 7.872711181640625, + "learning_rate": 1.1181367894096684e-06, + "logits/chosen": -2.3443212509155273, + "logits/rejected": -2.3754608631134033, + "logps/chosen": -25.16803550720215, + "logps/rejected": -14.663780212402344, + "loss": 0.8143, + "rewards/accuracies": 0.8125, + "rewards/chosen": 0.005207642912864685, + "rewards/margins": 1.0892767906188965, + "rewards/rejected": -1.0840692520141602, + "step": 656 + }, + { + "epoch": 1.442371020856202, + "grad_norm": 8.429574012756348, + "learning_rate": 1.1100668111113166e-06, + "logits/chosen": -2.3493621349334717, + "logits/rejected": -2.356614351272583, + "logps/chosen": -21.27967071533203, + "logps/rejected": -11.396280288696289, + "loss": 0.7689, + "rewards/accuracies": 0.78125, + "rewards/chosen": 0.11593085527420044, + "rewards/margins": 0.8025331497192383, + "rewards/rejected": -0.6866022348403931, + "step": 657 + }, + { + "epoch": 1.4445664105378704, + "grad_norm": 6.362731456756592, + "learning_rate": 1.1020177413231334e-06, + "logits/chosen": -2.336317300796509, + "logits/rejected": -2.393036365509033, + "logps/chosen": -21.837860107421875, + "logps/rejected": -11.513958930969238, + "loss": 0.8235, + "rewards/accuracies": 0.65625, + "rewards/chosen": 0.06914816796779633, + "rewards/margins": 0.8768153190612793, + "rewards/rejected": -0.8076671361923218, + "step": 658 + }, + { + "epoch": 1.446761800219539, + "grad_norm": 5.933378219604492, + "learning_rate": 1.0939897011258003e-06, + "logits/chosen": -2.3377718925476074, + "logits/rejected": -2.412820816040039, + "logps/chosen": -27.18308448791504, + "logps/rejected": -10.322875022888184, + "loss": 0.6968, + "rewards/accuracies": 0.78125, + "rewards/chosen": -0.010466039180755615, + "rewards/margins": 1.2958279848098755, + "rewards/rejected": -1.3062940835952759, + "step": 659 + }, + { + "epoch": 1.4489571899012075, + "grad_norm": 8.202812194824219, + "learning_rate": 1.085982811283654e-06, + "logits/chosen": -2.293506622314453, + "logits/rejected": -2.3609886169433594, + "logps/chosen": -25.594585418701172, + "logps/rejected": -11.83299446105957, + "loss": 0.8022, + "rewards/accuracies": 0.8125, + "rewards/chosen": -0.03281119093298912, + "rewards/margins": 1.2336405515670776, + "rewards/rejected": -1.2664518356323242, + "step": 660 + }, + { + "epoch": 1.451152579582876, + "grad_norm": 11.540955543518066, + "learning_rate": 1.077997192242871e-06, + "logits/chosen": -2.3301305770874023, + "logits/rejected": -2.3260154724121094, + "logps/chosen": -25.9770565032959, + "logps/rejected": -13.366264343261719, + "loss": 0.814, + "rewards/accuracies": 0.875, + "rewards/chosen": -0.016910400241613388, + "rewards/margins": 1.0012506246566772, + "rewards/rejected": -1.0181610584259033, + "step": 661 + }, + { + "epoch": 1.4533479692645446, + "grad_norm": 10.390963554382324, + "learning_rate": 1.0700329641296541e-06, + "logits/chosen": -2.361905097961426, + "logits/rejected": -2.3670222759246826, + "logps/chosen": -21.554336547851562, + "logps/rejected": -8.680063247680664, + "loss": 0.7527, + "rewards/accuracies": 0.71875, + "rewards/chosen": 0.009097723290324211, + "rewards/margins": 1.0622705221176147, + "rewards/rejected": -1.0531728267669678, + "step": 662 + }, + { + "epoch": 1.455543358946213, + "grad_norm": 7.903598785400391, + "learning_rate": 1.0620902467484279e-06, + "logits/chosen": -2.3902194499969482, + "logits/rejected": -2.3924710750579834, + "logps/chosen": -26.950511932373047, + "logps/rejected": -12.7181978225708, + "loss": 0.7504, + "rewards/accuracies": 0.875, + "rewards/chosen": 0.06110291928052902, + "rewards/margins": 1.0060982704162598, + "rewards/rejected": -0.9449952840805054, + "step": 663 + }, + { + "epoch": 1.4577387486278814, + "grad_norm": 6.64391565322876, + "learning_rate": 1.0541691595800338e-06, + "logits/chosen": -2.2780065536499023, + "logits/rejected": -2.2875936031341553, + "logps/chosen": -17.084646224975586, + "logps/rejected": -7.731725692749023, + "loss": 0.6781, + "rewards/accuracies": 0.78125, + "rewards/chosen": 0.11275922507047653, + "rewards/margins": 0.8099181652069092, + "rewards/rejected": -0.697158932685852, + "step": 664 + }, + { + "epoch": 1.4599341383095499, + "grad_norm": 10.162978172302246, + "learning_rate": 1.0462698217799333e-06, + "logits/chosen": -2.3904237747192383, + "logits/rejected": -2.334890365600586, + "logps/chosen": -19.63933563232422, + "logps/rejected": -13.295578956604004, + "loss": 0.8523, + "rewards/accuracies": 0.71875, + "rewards/chosen": 0.09234454482793808, + "rewards/margins": 0.8004003763198853, + "rewards/rejected": -0.708055853843689, + "step": 665 + }, + { + "epoch": 1.4621295279912183, + "grad_norm": 9.108477592468262, + "learning_rate": 1.0383923521764176e-06, + "logits/chosen": -2.338623523712158, + "logits/rejected": -2.2779643535614014, + "logps/chosen": -18.240127563476562, + "logps/rejected": -13.914999008178711, + "loss": 0.89, + "rewards/accuracies": 0.6875, + "rewards/chosen": 0.004548083059489727, + "rewards/margins": 0.8153023719787598, + "rewards/rejected": -0.8107542395591736, + "step": 666 + }, + { + "epoch": 1.464324917672887, + "grad_norm": 7.1527910232543945, + "learning_rate": 1.0305368692688175e-06, + "logits/chosen": -2.289700508117676, + "logits/rejected": -2.2619123458862305, + "logps/chosen": -21.673160552978516, + "logps/rejected": -9.784587860107422, + "loss": 0.7605, + "rewards/accuracies": 0.71875, + "rewards/chosen": 0.10032662749290466, + "rewards/margins": 1.0779885053634644, + "rewards/rejected": -0.9776619672775269, + "step": 667 + }, + { + "epoch": 1.4665203073545554, + "grad_norm": 10.545454025268555, + "learning_rate": 1.0227034912257222e-06, + "logits/chosen": -2.3333659172058105, + "logits/rejected": -2.3285748958587646, + "logps/chosen": -26.58244514465332, + "logps/rejected": -16.31827163696289, + "loss": 0.905, + "rewards/accuracies": 0.8125, + "rewards/chosen": -0.034804388880729675, + "rewards/margins": 1.0145269632339478, + "rewards/rejected": -1.0493313074111938, + "step": 668 + }, + { + "epoch": 1.4687156970362238, + "grad_norm": 10.760330200195312, + "learning_rate": 1.0148923358832022e-06, + "logits/chosen": -2.372616767883301, + "logits/rejected": -2.2945916652679443, + "logps/chosen": -19.306861877441406, + "logps/rejected": -10.68474006652832, + "loss": 0.745, + "rewards/accuracies": 0.78125, + "rewards/chosen": 0.051995545625686646, + "rewards/margins": 0.7446345686912537, + "rewards/rejected": -0.6926389932632446, + "step": 669 + }, + { + "epoch": 1.4709110867178925, + "grad_norm": 14.300837516784668, + "learning_rate": 1.0071035207430352e-06, + "logits/chosen": -2.4270987510681152, + "logits/rejected": -2.3773550987243652, + "logps/chosen": -27.409963607788086, + "logps/rejected": -12.27871036529541, + "loss": 0.7266, + "rewards/accuracies": 0.8125, + "rewards/chosen": 0.13977664709091187, + "rewards/margins": 1.289273977279663, + "rewards/rejected": -1.1494972705841064, + "step": 670 + }, + { + "epoch": 1.473106476399561, + "grad_norm": 8.178793907165527, + "learning_rate": 9.993371629709393e-07, + "logits/chosen": -2.2789974212646484, + "logits/rejected": -2.235015630722046, + "logps/chosen": -22.680187225341797, + "logps/rejected": -9.723089218139648, + "loss": 0.7926, + "rewards/accuracies": 0.65625, + "rewards/chosen": -0.018673136830329895, + "rewards/margins": 0.8131811618804932, + "rewards/rejected": -0.8318542242050171, + "step": 671 + }, + { + "epoch": 1.4753018660812294, + "grad_norm": 9.822272300720215, + "learning_rate": 9.91593379394811e-07, + "logits/chosen": -2.3281314373016357, + "logits/rejected": -2.3199617862701416, + "logps/chosen": -21.34670066833496, + "logps/rejected": -13.962909698486328, + "loss": 0.7335, + "rewards/accuracies": 0.90625, + "rewards/chosen": 0.15146546065807343, + "rewards/margins": 1.0111653804779053, + "rewards/rejected": -0.8596999049186707, + "step": 672 + }, + { + "epoch": 1.477497255762898, + "grad_norm": 5.761569976806641, + "learning_rate": 9.838722865029674e-07, + "logits/chosen": -2.2526254653930664, + "logits/rejected": -2.3039088249206543, + "logps/chosen": -27.100936889648438, + "logps/rejected": -11.220876693725586, + "loss": 0.7244, + "rewards/accuracies": 0.84375, + "rewards/chosen": 0.10972346365451813, + "rewards/margins": 1.18520188331604, + "rewards/rejected": -1.0754785537719727, + "step": 673 + }, + { + "epoch": 1.4796926454445665, + "grad_norm": 8.87966537475586, + "learning_rate": 9.761740004423928e-07, + "logits/chosen": -2.3781142234802246, + "logits/rejected": -2.2708754539489746, + "logps/chosen": -28.19532012939453, + "logps/rejected": -10.572488784790039, + "loss": 0.6689, + "rewards/accuracies": 0.8125, + "rewards/chosen": -0.01809457316994667, + "rewards/margins": 1.1581518650054932, + "rewards/rejected": -1.1762464046478271, + "step": 674 + }, + { + "epoch": 1.481888035126235, + "grad_norm": 7.108104705810547, + "learning_rate": 9.68498637016993e-07, + "logits/chosen": -2.2991137504577637, + "logits/rejected": -2.289788246154785, + "logps/chosen": -20.59160041809082, + "logps/rejected": -10.33639144897461, + "loss": 0.7128, + "rewards/accuracies": 0.875, + "rewards/chosen": 0.0434945784509182, + "rewards/margins": 1.0837714672088623, + "rewards/rejected": -1.0402767658233643, + "step": 675 + }, + { + "epoch": 1.4840834248079033, + "grad_norm": 9.927098274230957, + "learning_rate": 9.608463116858544e-07, + "logits/chosen": -2.3607635498046875, + "logits/rejected": -2.3653457164764404, + "logps/chosen": -23.1501522064209, + "logps/rejected": -15.188650131225586, + "loss": 0.986, + "rewards/accuracies": 0.65625, + "rewards/chosen": -0.08851507306098938, + "rewards/margins": 0.6813887357711792, + "rewards/rejected": -0.7699037790298462, + "step": 676 + }, + { + "epoch": 1.4862788144895718, + "grad_norm": 10.931567192077637, + "learning_rate": 9.532171395615036e-07, + "logits/chosen": -2.2876527309417725, + "logits/rejected": -2.3687150478363037, + "logps/chosen": -23.15837860107422, + "logps/rejected": -11.48902416229248, + "loss": 0.7481, + "rewards/accuracies": 0.75, + "rewards/chosen": -0.02260664477944374, + "rewards/margins": 0.9332041144371033, + "rewards/rejected": -0.9558107852935791, + "step": 677 + }, + { + "epoch": 1.4884742041712404, + "grad_norm": 7.135960578918457, + "learning_rate": 9.456112354081781e-07, + "logits/chosen": -2.3467886447906494, + "logits/rejected": -2.436093330383301, + "logps/chosen": -25.84896469116211, + "logps/rejected": -12.919902801513672, + "loss": 0.815, + "rewards/accuracies": 0.875, + "rewards/chosen": 0.029841048642992973, + "rewards/margins": 1.0429695844650269, + "rewards/rejected": -1.0131285190582275, + "step": 678 + }, + { + "epoch": 1.4906695938529089, + "grad_norm": 7.798811912536621, + "learning_rate": 9.380287136401001e-07, + "logits/chosen": -2.290736436843872, + "logits/rejected": -2.3213000297546387, + "logps/chosen": -21.170406341552734, + "logps/rejected": -11.270086288452148, + "loss": 0.8228, + "rewards/accuracies": 0.71875, + "rewards/chosen": 0.03362353518605232, + "rewards/margins": 0.669628381729126, + "rewards/rejected": -0.6360048651695251, + "step": 679 + }, + { + "epoch": 1.4928649835345773, + "grad_norm": 4.894595623016357, + "learning_rate": 9.304696883197542e-07, + "logits/chosen": -2.287060499191284, + "logits/rejected": -2.396371841430664, + "logps/chosen": -23.43198013305664, + "logps/rejected": -10.551751136779785, + "loss": 0.7583, + "rewards/accuracies": 0.875, + "rewards/chosen": 0.07391675561666489, + "rewards/margins": 1.0375158786773682, + "rewards/rejected": -0.9635990858078003, + "step": 680 + }, + { + "epoch": 1.495060373216246, + "grad_norm": 10.179999351501465, + "learning_rate": 9.229342731561719e-07, + "logits/chosen": -2.329038143157959, + "logits/rejected": -2.305812358856201, + "logps/chosen": -24.562477111816406, + "logps/rejected": -14.272148132324219, + "loss": 0.8547, + "rewards/accuracies": 0.65625, + "rewards/chosen": 0.003495384007692337, + "rewards/margins": 0.8332772254943848, + "rewards/rejected": -0.8297817707061768, + "step": 681 + }, + { + "epoch": 1.4972557628979144, + "grad_norm": 5.932125091552734, + "learning_rate": 9.154225815032242e-07, + "logits/chosen": -2.2871360778808594, + "logits/rejected": -2.29762864112854, + "logps/chosen": -25.172122955322266, + "logps/rejected": -11.337629318237305, + "loss": 0.7278, + "rewards/accuracies": 0.8125, + "rewards/chosen": -0.004267069511115551, + "rewards/margins": 0.9601096510887146, + "rewards/rejected": -0.96437668800354, + "step": 682 + }, + { + "epoch": 1.4994511525795828, + "grad_norm": 7.449439525604248, + "learning_rate": 9.079347263579103e-07, + "logits/chosen": -2.402919292449951, + "logits/rejected": -2.342090368270874, + "logps/chosen": -21.132667541503906, + "logps/rejected": -10.882007598876953, + "loss": 0.7609, + "rewards/accuracies": 0.8125, + "rewards/chosen": 0.044458843767642975, + "rewards/margins": 0.910749614238739, + "rewards/rejected": -0.8662907481193542, + "step": 683 + }, + { + "epoch": 1.5016465422612515, + "grad_norm": 10.604455947875977, + "learning_rate": 9.004708203586629e-07, + "logits/chosen": -2.3037948608398438, + "logits/rejected": -2.3292269706726074, + "logps/chosen": -20.89158058166504, + "logps/rejected": -12.445920944213867, + "loss": 0.7671, + "rewards/accuracies": 0.8125, + "rewards/chosen": 0.022653302177786827, + "rewards/margins": 0.8269745111465454, + "rewards/rejected": -0.8043211698532104, + "step": 684 + }, + { + "epoch": 1.5038419319429197, + "grad_norm": 8.496170043945312, + "learning_rate": 8.930309757836517e-07, + "logits/chosen": -2.290376901626587, + "logits/rejected": -2.3371200561523438, + "logps/chosen": -26.557998657226562, + "logps/rejected": -16.33134651184082, + "loss": 0.8505, + "rewards/accuracies": 0.78125, + "rewards/chosen": 0.1173873245716095, + "rewards/margins": 1.1820646524429321, + "rewards/rejected": -1.0646772384643555, + "step": 685 + }, + { + "epoch": 1.5060373216245884, + "grad_norm": 7.814933776855469, + "learning_rate": 8.856153045490948e-07, + "logits/chosen": -2.373056411743164, + "logits/rejected": -2.377485990524292, + "logps/chosen": -21.816408157348633, + "logps/rejected": -7.910917282104492, + "loss": 0.6258, + "rewards/accuracies": 0.875, + "rewards/chosen": 0.13469111919403076, + "rewards/margins": 1.124887228012085, + "rewards/rejected": -0.9901960492134094, + "step": 686 + }, + { + "epoch": 1.508232711306257, + "grad_norm": 10.497405052185059, + "learning_rate": 8.78223918207575e-07, + "logits/chosen": -2.330064535140991, + "logits/rejected": -2.4031729698181152, + "logps/chosen": -24.309425354003906, + "logps/rejected": -11.734188079833984, + "loss": 0.7726, + "rewards/accuracies": 0.875, + "rewards/chosen": 0.06004462391138077, + "rewards/margins": 0.9490844011306763, + "rewards/rejected": -0.8890397548675537, + "step": 687 + }, + { + "epoch": 1.5104281009879252, + "grad_norm": 7.284695625305176, + "learning_rate": 8.708569279463622e-07, + "logits/chosen": -2.3188695907592773, + "logits/rejected": -2.31721568107605, + "logps/chosen": -18.19530487060547, + "logps/rejected": -10.189314842224121, + "loss": 0.7654, + "rewards/accuracies": 0.8125, + "rewards/chosen": 0.0393633097410202, + "rewards/margins": 0.8210750818252563, + "rewards/rejected": -0.7817118167877197, + "step": 688 + }, + { + "epoch": 1.512623490669594, + "grad_norm": 9.467089653015137, + "learning_rate": 8.635144445857407e-07, + "logits/chosen": -2.3692595958709717, + "logits/rejected": -2.367439031600952, + "logps/chosen": -23.300127029418945, + "logps/rejected": -9.973893165588379, + "loss": 0.6943, + "rewards/accuracies": 0.8125, + "rewards/chosen": 0.012691512703895569, + "rewards/margins": 1.2339590787887573, + "rewards/rejected": -1.221267580986023, + "step": 689 + }, + { + "epoch": 1.5148188803512623, + "grad_norm": 6.377358436584473, + "learning_rate": 8.561965785773413e-07, + "logits/chosen": -2.3550801277160645, + "logits/rejected": -2.2900638580322266, + "logps/chosen": -21.67240333557129, + "logps/rejected": -11.367239952087402, + "loss": 0.7081, + "rewards/accuracies": 0.96875, + "rewards/chosen": 0.10215982794761658, + "rewards/margins": 1.0111408233642578, + "rewards/rejected": -0.9089809656143188, + "step": 690 + }, + { + "epoch": 1.5170142700329308, + "grad_norm": 6.283344268798828, + "learning_rate": 8.489034400024812e-07, + "logits/chosen": -2.311403274536133, + "logits/rejected": -2.2807555198669434, + "logps/chosen": -27.165481567382812, + "logps/rejected": -16.081340789794922, + "loss": 0.9181, + "rewards/accuracies": 0.90625, + "rewards/chosen": 0.023088926449418068, + "rewards/margins": 0.9116746187210083, + "rewards/rejected": -0.8885857462882996, + "step": 691 + }, + { + "epoch": 1.5192096597145994, + "grad_norm": 6.94219970703125, + "learning_rate": 8.416351385705071e-07, + "logits/chosen": -2.2564587593078613, + "logits/rejected": -2.270942211151123, + "logps/chosen": -19.77495765686035, + "logps/rejected": -11.380971908569336, + "loss": 0.7348, + "rewards/accuracies": 0.875, + "rewards/chosen": 0.00265529565513134, + "rewards/margins": 0.9093756675720215, + "rewards/rejected": -0.9067203402519226, + "step": 692 + }, + { + "epoch": 1.5214050493962679, + "grad_norm": 6.930017948150635, + "learning_rate": 8.343917836171448e-07, + "logits/chosen": -2.3904213905334473, + "logits/rejected": -2.3949031829833984, + "logps/chosen": -20.12192153930664, + "logps/rejected": -13.20472526550293, + "loss": 0.873, + "rewards/accuracies": 0.71875, + "rewards/chosen": -0.008965975604951382, + "rewards/margins": 0.7542557120323181, + "rewards/rejected": -0.7632217407226562, + "step": 693 + }, + { + "epoch": 1.5236004390779363, + "grad_norm": 11.935271263122559, + "learning_rate": 8.271734841028553e-07, + "logits/chosen": -2.395549774169922, + "logits/rejected": -2.3498220443725586, + "logps/chosen": -20.41801643371582, + "logps/rejected": -13.415026664733887, + "loss": 0.8859, + "rewards/accuracies": 0.75, + "rewards/chosen": -0.047664545476436615, + "rewards/margins": 0.8935979604721069, + "rewards/rejected": -0.9412623643875122, + "step": 694 + }, + { + "epoch": 1.525795828759605, + "grad_norm": 6.534700870513916, + "learning_rate": 8.19980348611194e-07, + "logits/chosen": -2.2631173133850098, + "logits/rejected": -2.331083059310913, + "logps/chosen": -19.663105010986328, + "logps/rejected": -10.302260398864746, + "loss": 0.7654, + "rewards/accuracies": 0.71875, + "rewards/chosen": 0.12376576662063599, + "rewards/margins": 0.8773860931396484, + "rewards/rejected": -0.7536203861236572, + "step": 695 + }, + { + "epoch": 1.5279912184412732, + "grad_norm": 6.508988380432129, + "learning_rate": 8.128124853471814e-07, + "logits/chosen": -2.3251452445983887, + "logits/rejected": -2.284334659576416, + "logps/chosen": -19.876367568969727, + "logps/rejected": -11.618218421936035, + "loss": 0.8347, + "rewards/accuracies": 0.71875, + "rewards/chosen": 0.07319276034832001, + "rewards/margins": 0.6937046051025391, + "rewards/rejected": -0.6205118298530579, + "step": 696 + }, + { + "epoch": 1.5301866081229418, + "grad_norm": 10.9249267578125, + "learning_rate": 8.056700021356695e-07, + "logits/chosen": -2.3364064693450928, + "logits/rejected": -2.2750191688537598, + "logps/chosen": -24.767620086669922, + "logps/rejected": -12.226917266845703, + "loss": 0.7793, + "rewards/accuracies": 0.875, + "rewards/chosen": -0.04602807015180588, + "rewards/margins": 0.9939512610435486, + "rewards/rejected": -1.039979338645935, + "step": 697 + }, + { + "epoch": 1.5323819978046103, + "grad_norm": 9.46987247467041, + "learning_rate": 7.985530064197242e-07, + "logits/chosen": -2.3092966079711914, + "logits/rejected": -2.3300094604492188, + "logps/chosen": -21.44502067565918, + "logps/rejected": -9.953096389770508, + "loss": 0.7574, + "rewards/accuracies": 0.8125, + "rewards/chosen": -0.005251489579677582, + "rewards/margins": 0.6864630579948425, + "rewards/rejected": -0.6917145848274231, + "step": 698 + }, + { + "epoch": 1.5345773874862787, + "grad_norm": 11.609639167785645, + "learning_rate": 7.914616052590071e-07, + "logits/chosen": -2.2799625396728516, + "logits/rejected": -2.376351833343506, + "logps/chosen": -20.676040649414062, + "logps/rejected": -15.289880752563477, + "loss": 0.9793, + "rewards/accuracies": 0.6875, + "rewards/chosen": -0.07748284190893173, + "rewards/margins": 0.663689136505127, + "rewards/rejected": -0.7411720156669617, + "step": 699 + }, + { + "epoch": 1.5367727771679474, + "grad_norm": 9.140620231628418, + "learning_rate": 7.843959053281663e-07, + "logits/chosen": -2.3910202980041504, + "logits/rejected": -2.3386998176574707, + "logps/chosen": -20.352893829345703, + "logps/rejected": -11.073060989379883, + "loss": 0.7694, + "rewards/accuracies": 0.75, + "rewards/chosen": 0.0445685088634491, + "rewards/margins": 0.8274343013763428, + "rewards/rejected": -0.7828657627105713, + "step": 700 + }, + { + "epoch": 1.5389681668496158, + "grad_norm": 6.100501537322998, + "learning_rate": 7.7735601291523e-07, + "logits/chosen": -2.2963991165161133, + "logits/rejected": -2.307669162750244, + "logps/chosen": -20.801546096801758, + "logps/rejected": -10.02623176574707, + "loss": 0.7519, + "rewards/accuracies": 0.8125, + "rewards/chosen": 0.10802704095840454, + "rewards/margins": 0.9469428062438965, + "rewards/rejected": -0.8389157652854919, + "step": 701 + }, + { + "epoch": 1.5411635565312842, + "grad_norm": 7.051187515258789, + "learning_rate": 7.703420339200102e-07, + "logits/chosen": -2.3196403980255127, + "logits/rejected": -2.4223508834838867, + "logps/chosen": -24.035518646240234, + "logps/rejected": -10.482366561889648, + "loss": 0.7986, + "rewards/accuracies": 0.65625, + "rewards/chosen": -0.02300884947180748, + "rewards/margins": 1.0546743869781494, + "rewards/rejected": -1.0776833295822144, + "step": 702 + }, + { + "epoch": 1.5433589462129529, + "grad_norm": 5.677674770355225, + "learning_rate": 7.633540738525066e-07, + "logits/chosen": -2.3929460048675537, + "logits/rejected": -2.3577888011932373, + "logps/chosen": -24.904727935791016, + "logps/rejected": -9.811440467834473, + "loss": 0.6212, + "rewards/accuracies": 0.84375, + "rewards/chosen": 0.18078622221946716, + "rewards/margins": 1.3551355600357056, + "rewards/rejected": -1.174349308013916, + "step": 703 + }, + { + "epoch": 1.5455543358946213, + "grad_norm": 7.708553791046143, + "learning_rate": 7.563922378313218e-07, + "logits/chosen": -2.286649227142334, + "logits/rejected": -2.342291831970215, + "logps/chosen": -22.24799346923828, + "logps/rejected": -10.20236587524414, + "loss": 0.7709, + "rewards/accuracies": 0.75, + "rewards/chosen": -0.004813693463802338, + "rewards/margins": 0.9623636603355408, + "rewards/rejected": -0.9671773910522461, + "step": 704 + }, + { + "epoch": 1.5477497255762898, + "grad_norm": 10.207706451416016, + "learning_rate": 7.494566305820788e-07, + "logits/chosen": -2.3124635219573975, + "logits/rejected": -2.3873467445373535, + "logps/chosen": -22.17903709411621, + "logps/rejected": -11.572494506835938, + "loss": 0.8085, + "rewards/accuracies": 0.78125, + "rewards/chosen": 0.060527339577674866, + "rewards/margins": 0.9175864458084106, + "rewards/rejected": -0.8570590615272522, + "step": 705 + }, + { + "epoch": 1.5499451152579584, + "grad_norm": 9.1857271194458, + "learning_rate": 7.425473564358457e-07, + "logits/chosen": -2.279327630996704, + "logits/rejected": -2.299070358276367, + "logps/chosen": -23.58930206298828, + "logps/rejected": -10.550365447998047, + "loss": 0.7257, + "rewards/accuracies": 0.84375, + "rewards/chosen": 0.05851493030786514, + "rewards/margins": 1.1829012632369995, + "rewards/rejected": -1.1243863105773926, + "step": 706 + }, + { + "epoch": 1.5521405049396266, + "grad_norm": 14.050537109375, + "learning_rate": 7.356645193275675e-07, + "logits/chosen": -2.3987393379211426, + "logits/rejected": -2.372545003890991, + "logps/chosen": -25.122947692871094, + "logps/rejected": -15.814371109008789, + "loss": 0.825, + "rewards/accuracies": 0.90625, + "rewards/chosen": 0.04791188985109329, + "rewards/margins": 1.3396070003509521, + "rewards/rejected": -1.2916948795318604, + "step": 707 + }, + { + "epoch": 1.5543358946212953, + "grad_norm": 8.37563419342041, + "learning_rate": 7.288082227945007e-07, + "logits/chosen": -2.355762243270874, + "logits/rejected": -2.291193962097168, + "logps/chosen": -17.685192108154297, + "logps/rejected": -9.716022491455078, + "loss": 0.7535, + "rewards/accuracies": 0.8125, + "rewards/chosen": 0.06530803442001343, + "rewards/margins": 0.6777411699295044, + "rewards/rejected": -0.6124331951141357, + "step": 708 + }, + { + "epoch": 1.5565312843029637, + "grad_norm": 6.0413994789123535, + "learning_rate": 7.219785699746573e-07, + "logits/chosen": -2.2274327278137207, + "logits/rejected": -2.287482738494873, + "logps/chosen": -23.955791473388672, + "logps/rejected": -8.871516227722168, + "loss": 0.6564, + "rewards/accuracies": 0.84375, + "rewards/chosen": 0.07103793323040009, + "rewards/margins": 1.381327509880066, + "rewards/rejected": -1.3102896213531494, + "step": 709 + }, + { + "epoch": 1.5587266739846322, + "grad_norm": 12.714701652526855, + "learning_rate": 7.151756636052529e-07, + "logits/chosen": -2.299858331680298, + "logits/rejected": -2.387424945831299, + "logps/chosen": -19.79357147216797, + "logps/rejected": -9.718650817871094, + "loss": 0.7547, + "rewards/accuracies": 0.6875, + "rewards/chosen": 0.02891990728676319, + "rewards/margins": 0.8539602756500244, + "rewards/rejected": -0.8250404000282288, + "step": 710 + }, + { + "epoch": 1.5609220636663008, + "grad_norm": 6.89093542098999, + "learning_rate": 7.083996060211606e-07, + "logits/chosen": -2.2775955200195312, + "logits/rejected": -2.2080445289611816, + "logps/chosen": -22.861404418945312, + "logps/rejected": -12.659036636352539, + "loss": 0.8498, + "rewards/accuracies": 0.8125, + "rewards/chosen": 0.07481466233730316, + "rewards/margins": 0.6025862693786621, + "rewards/rejected": -0.5277715921401978, + "step": 711 + }, + { + "epoch": 1.5631174533479693, + "grad_norm": 7.596837520599365, + "learning_rate": 7.016504991533727e-07, + "logits/chosen": -2.3061470985412598, + "logits/rejected": -2.308680534362793, + "logps/chosen": -22.837581634521484, + "logps/rejected": -12.386337280273438, + "loss": 0.7578, + "rewards/accuracies": 0.78125, + "rewards/chosen": 0.13396987318992615, + "rewards/margins": 1.1027204990386963, + "rewards/rejected": -0.968750536441803, + "step": 712 + }, + { + "epoch": 1.5653128430296377, + "grad_norm": 11.662274360656738, + "learning_rate": 6.949284445274667e-07, + "logits/chosen": -2.267864227294922, + "logits/rejected": -2.2811851501464844, + "logps/chosen": -26.040233612060547, + "logps/rejected": -14.082418441772461, + "loss": 0.7907, + "rewards/accuracies": 0.84375, + "rewards/chosen": 0.029413022100925446, + "rewards/margins": 1.2451152801513672, + "rewards/rejected": -1.2157021760940552, + "step": 713 + }, + { + "epoch": 1.5675082327113063, + "grad_norm": 7.059804439544678, + "learning_rate": 6.882335432620779e-07, + "logits/chosen": -2.3052608966827393, + "logits/rejected": -2.362246036529541, + "logps/chosen": -25.100984573364258, + "logps/rejected": -12.064314842224121, + "loss": 0.7653, + "rewards/accuracies": 0.78125, + "rewards/chosen": 0.035489220172166824, + "rewards/margins": 1.2769728899002075, + "rewards/rejected": -1.2414836883544922, + "step": 714 + }, + { + "epoch": 1.5697036223929748, + "grad_norm": 9.266624450683594, + "learning_rate": 6.815658960673782e-07, + "logits/chosen": -2.3363451957702637, + "logits/rejected": -2.236959934234619, + "logps/chosen": -20.818782806396484, + "logps/rejected": -12.962738037109375, + "loss": 0.8648, + "rewards/accuracies": 0.6875, + "rewards/chosen": 0.049082882702350616, + "rewards/margins": 0.7642035484313965, + "rewards/rejected": -0.7151206731796265, + "step": 715 + }, + { + "epoch": 1.5718990120746432, + "grad_norm": 11.131867408752441, + "learning_rate": 6.749256032435636e-07, + "logits/chosen": -2.2627530097961426, + "logits/rejected": -2.346843957901001, + "logps/chosen": -25.85169219970703, + "logps/rejected": -11.823819160461426, + "loss": 0.8033, + "rewards/accuracies": 0.625, + "rewards/chosen": 0.061939992010593414, + "rewards/margins": 1.1473208665847778, + "rewards/rejected": -1.0853807926177979, + "step": 716 + }, + { + "epoch": 1.5740944017563119, + "grad_norm": 6.476619243621826, + "learning_rate": 6.683127646793411e-07, + "logits/chosen": -2.335946559906006, + "logits/rejected": -2.379199743270874, + "logps/chosen": -23.626924514770508, + "logps/rejected": -10.17281723022461, + "loss": 0.772, + "rewards/accuracies": 0.78125, + "rewards/chosen": 0.13880592584609985, + "rewards/margins": 0.8892215490341187, + "rewards/rejected": -0.7504156827926636, + "step": 717 + }, + { + "epoch": 1.57628979143798, + "grad_norm": 13.723481178283691, + "learning_rate": 6.617274798504286e-07, + "logits/chosen": -2.4252548217773438, + "logits/rejected": -2.4306135177612305, + "logps/chosen": -18.858985900878906, + "logps/rejected": -12.96285629272461, + "loss": 0.8827, + "rewards/accuracies": 0.71875, + "rewards/chosen": -0.09428222477436066, + "rewards/margins": 0.654495358467102, + "rewards/rejected": -0.7487775087356567, + "step": 718 + }, + { + "epoch": 1.5784851811196488, + "grad_norm": 8.539143562316895, + "learning_rate": 6.55169847818059e-07, + "logits/chosen": -2.307605743408203, + "logits/rejected": -2.29728364944458, + "logps/chosen": -21.42572021484375, + "logps/rejected": -8.90909194946289, + "loss": 0.6328, + "rewards/accuracies": 0.875, + "rewards/chosen": 0.13041691482067108, + "rewards/margins": 1.1438201665878296, + "rewards/rejected": -1.013403296470642, + "step": 719 + }, + { + "epoch": 1.5806805708013172, + "grad_norm": 7.203423023223877, + "learning_rate": 6.48639967227489e-07, + "logits/chosen": -2.3547959327697754, + "logits/rejected": -2.433210849761963, + "logps/chosen": -25.69765281677246, + "logps/rejected": -9.445756912231445, + "loss": 0.6243, + "rewards/accuracies": 0.90625, + "rewards/chosen": 0.13975399732589722, + "rewards/margins": 1.5468833446502686, + "rewards/rejected": -1.4071294069290161, + "step": 720 + }, + { + "epoch": 1.5828759604829856, + "grad_norm": 6.584375381469727, + "learning_rate": 6.421379363065142e-07, + "logits/chosen": -2.3256547451019287, + "logits/rejected": -2.2872846126556396, + "logps/chosen": -22.967172622680664, + "logps/rejected": -13.148837089538574, + "loss": 0.7659, + "rewards/accuracies": 0.84375, + "rewards/chosen": 0.12208892405033112, + "rewards/margins": 1.055436134338379, + "rewards/rejected": -0.9333471059799194, + "step": 721 + }, + { + "epoch": 1.5850713501646543, + "grad_norm": 8.888541221618652, + "learning_rate": 6.356638528639955e-07, + "logits/chosen": -2.2938733100891113, + "logits/rejected": -2.327450752258301, + "logps/chosen": -20.15079116821289, + "logps/rejected": -10.908607482910156, + "loss": 0.777, + "rewards/accuracies": 0.75, + "rewards/chosen": 0.09680142253637314, + "rewards/margins": 0.822465181350708, + "rewards/rejected": -0.7256637811660767, + "step": 722 + }, + { + "epoch": 1.5872667398463227, + "grad_norm": 6.514217853546143, + "learning_rate": 6.292178142883826e-07, + "logits/chosen": -2.254856586456299, + "logits/rejected": -2.340163469314575, + "logps/chosen": -21.568584442138672, + "logps/rejected": -13.060552597045898, + "loss": 0.8386, + "rewards/accuracies": 0.75, + "rewards/chosen": 0.13580599427223206, + "rewards/margins": 0.6412376165390015, + "rewards/rejected": -0.5054316520690918, + "step": 723 + }, + { + "epoch": 1.5894621295279912, + "grad_norm": 7.364799976348877, + "learning_rate": 6.227999175462521e-07, + "logits/chosen": -2.318565607070923, + "logits/rejected": -2.3171820640563965, + "logps/chosen": -23.852569580078125, + "logps/rejected": -9.777791023254395, + "loss": 0.694, + "rewards/accuracies": 0.90625, + "rewards/chosen": 0.06606684625148773, + "rewards/margins": 1.2393674850463867, + "rewards/rejected": -1.1733007431030273, + "step": 724 + }, + { + "epoch": 1.5916575192096598, + "grad_norm": 7.290419578552246, + "learning_rate": 6.164102591808482e-07, + "logits/chosen": -2.3687431812286377, + "logits/rejected": -2.257507801055908, + "logps/chosen": -21.314414978027344, + "logps/rejected": -12.256153106689453, + "loss": 0.7955, + "rewards/accuracies": 0.84375, + "rewards/chosen": 0.029902014881372452, + "rewards/margins": 0.9053773283958435, + "rewards/rejected": -0.8754753470420837, + "step": 725 + }, + { + "epoch": 1.5938529088913282, + "grad_norm": 5.709447860717773, + "learning_rate": 6.100489353106303e-07, + "logits/chosen": -2.225846290588379, + "logits/rejected": -2.2001028060913086, + "logps/chosen": -21.839332580566406, + "logps/rejected": -10.839560508728027, + "loss": 0.7676, + "rewards/accuracies": 0.90625, + "rewards/chosen": -0.01533023826777935, + "rewards/margins": 0.8626635074615479, + "rewards/rejected": -0.8779938220977783, + "step": 726 + }, + { + "epoch": 1.5960482985729967, + "grad_norm": 7.302804946899414, + "learning_rate": 6.037160416278279e-07, + "logits/chosen": -2.3346877098083496, + "logits/rejected": -2.245436429977417, + "logps/chosen": -24.075117111206055, + "logps/rejected": -8.489334106445312, + "loss": 0.5795, + "rewards/accuracies": 1.0, + "rewards/chosen": 0.19620266556739807, + "rewards/margins": 1.303248405456543, + "rewards/rejected": -1.1070456504821777, + "step": 727 + }, + { + "epoch": 1.5982436882546653, + "grad_norm": 5.894931316375732, + "learning_rate": 5.974116733969989e-07, + "logits/chosen": -2.3605470657348633, + "logits/rejected": -2.2782344818115234, + "logps/chosen": -21.21055793762207, + "logps/rejected": -13.394701957702637, + "loss": 0.7951, + "rewards/accuracies": 0.71875, + "rewards/chosen": 0.01902463473379612, + "rewards/margins": 0.8854861259460449, + "rewards/rejected": -0.8664615154266357, + "step": 728 + }, + { + "epoch": 1.6004390779363336, + "grad_norm": 6.547585964202881, + "learning_rate": 5.911359254535995e-07, + "logits/chosen": -2.3427815437316895, + "logits/rejected": -2.3127663135528564, + "logps/chosen": -24.723941802978516, + "logps/rejected": -14.315020561218262, + "loss": 0.8718, + "rewards/accuracies": 0.71875, + "rewards/chosen": -0.10273928940296173, + "rewards/margins": 0.7579648494720459, + "rewards/rejected": -0.8607041835784912, + "step": 729 + }, + { + "epoch": 1.6026344676180022, + "grad_norm": 8.056953430175781, + "learning_rate": 5.848888922025553e-07, + "logits/chosen": -2.230651378631592, + "logits/rejected": -2.2704825401306152, + "logps/chosen": -22.11258316040039, + "logps/rejected": -15.916872024536133, + "loss": 0.8679, + "rewards/accuracies": 0.78125, + "rewards/chosen": 0.03399306908249855, + "rewards/margins": 1.0302367210388184, + "rewards/rejected": -0.9962435960769653, + "step": 730 + }, + { + "epoch": 1.6048298572996706, + "grad_norm": 6.528773784637451, + "learning_rate": 5.786706676168424e-07, + "logits/chosen": -2.2501115798950195, + "logits/rejected": -2.286118507385254, + "logps/chosen": -22.918170928955078, + "logps/rejected": -9.76865291595459, + "loss": 0.6636, + "rewards/accuracies": 0.875, + "rewards/chosen": 0.09707888960838318, + "rewards/margins": 1.059900164604187, + "rewards/rejected": -0.962821364402771, + "step": 731 + }, + { + "epoch": 1.607025246981339, + "grad_norm": 6.323126316070557, + "learning_rate": 5.724813452360736e-07, + "logits/chosen": -2.3105685710906982, + "logits/rejected": -2.4109482765197754, + "logps/chosen": -22.616268157958984, + "logps/rejected": -11.512091636657715, + "loss": 0.7844, + "rewards/accuracies": 0.6875, + "rewards/chosen": 0.06783854961395264, + "rewards/margins": 0.9781845808029175, + "rewards/rejected": -0.9103460311889648, + "step": 732 + }, + { + "epoch": 1.6092206366630077, + "grad_norm": 8.920979499816895, + "learning_rate": 5.663210181650905e-07, + "logits/chosen": -2.2927749156951904, + "logits/rejected": -2.3700671195983887, + "logps/chosen": -32.030052185058594, + "logps/rejected": -15.700666427612305, + "loss": 0.7788, + "rewards/accuracies": 0.8125, + "rewards/chosen": -0.11107874661684036, + "rewards/margins": 1.1730841398239136, + "rewards/rejected": -1.284162998199463, + "step": 733 + }, + { + "epoch": 1.6114160263446762, + "grad_norm": 7.94144344329834, + "learning_rate": 5.601897790725644e-07, + "logits/chosen": -2.2817940711975098, + "logits/rejected": -2.362081289291382, + "logps/chosen": -26.45559310913086, + "logps/rejected": -11.353103637695312, + "loss": 0.7284, + "rewards/accuracies": 0.875, + "rewards/chosen": 0.13646835088729858, + "rewards/margins": 1.4305644035339355, + "rewards/rejected": -1.2940959930419922, + "step": 734 + }, + { + "epoch": 1.6136114160263446, + "grad_norm": 10.363531112670898, + "learning_rate": 5.540877201896e-07, + "logits/chosen": -2.3893866539001465, + "logits/rejected": -2.2922868728637695, + "logps/chosen": -19.933313369750977, + "logps/rejected": -11.359781265258789, + "loss": 0.8002, + "rewards/accuracies": 0.84375, + "rewards/chosen": 0.00043469667434692383, + "rewards/margins": 0.7254297733306885, + "rewards/rejected": -0.7249950170516968, + "step": 735 + }, + { + "epoch": 1.6158068057080133, + "grad_norm": 7.4115495681762695, + "learning_rate": 5.48014933308352e-07, + "logits/chosen": -2.3513894081115723, + "logits/rejected": -2.326012372970581, + "logps/chosen": -22.900840759277344, + "logps/rejected": -11.856263160705566, + "loss": 0.6798, + "rewards/accuracies": 0.84375, + "rewards/chosen": 0.06937343627214432, + "rewards/margins": 1.2832187414169312, + "rewards/rejected": -1.2138452529907227, + "step": 736 + }, + { + "epoch": 1.6180021953896817, + "grad_norm": 10.967841148376465, + "learning_rate": 5.419715097806397e-07, + "logits/chosen": -2.317211627960205, + "logits/rejected": -2.3183681964874268, + "logps/chosen": -21.147233963012695, + "logps/rejected": -11.363478660583496, + "loss": 0.7847, + "rewards/accuracies": 0.84375, + "rewards/chosen": 0.07483860105276108, + "rewards/margins": 0.773987889289856, + "rewards/rejected": -0.6991493105888367, + "step": 737 + }, + { + "epoch": 1.6201975850713501, + "grad_norm": 6.511829376220703, + "learning_rate": 5.359575405165749e-07, + "logits/chosen": -2.321399211883545, + "logits/rejected": -2.3429579734802246, + "logps/chosen": -26.204055786132812, + "logps/rejected": -10.985153198242188, + "loss": 0.7253, + "rewards/accuracies": 0.8125, + "rewards/chosen": -0.059297651052474976, + "rewards/margins": 1.0806995630264282, + "rewards/rejected": -1.1399972438812256, + "step": 738 + }, + { + "epoch": 1.6223929747530188, + "grad_norm": 6.037580966949463, + "learning_rate": 5.299731159831953e-07, + "logits/chosen": -2.2761242389678955, + "logits/rejected": -2.298978090286255, + "logps/chosen": -24.696460723876953, + "logps/rejected": -11.650628089904785, + "loss": 0.7296, + "rewards/accuracies": 0.84375, + "rewards/chosen": 0.03301388397812843, + "rewards/margins": 1.119810938835144, + "rewards/rejected": -1.0867971181869507, + "step": 739 + }, + { + "epoch": 1.624588364434687, + "grad_norm": 5.7746052742004395, + "learning_rate": 5.240183262031021e-07, + "logits/chosen": -2.264317750930786, + "logits/rejected": -2.257780075073242, + "logps/chosen": -19.13739013671875, + "logps/rejected": -7.076645851135254, + "loss": 0.644, + "rewards/accuracies": 0.875, + "rewards/chosen": 0.16118277609348297, + "rewards/margins": 0.9924280643463135, + "rewards/rejected": -0.8312452435493469, + "step": 740 + }, + { + "epoch": 1.6267837541163557, + "grad_norm": 10.51162052154541, + "learning_rate": 5.180932607531055e-07, + "logits/chosen": -2.326261281967163, + "logits/rejected": -2.28745698928833, + "logps/chosen": -20.67829132080078, + "logps/rejected": -13.597274780273438, + "loss": 0.8564, + "rewards/accuracies": 0.78125, + "rewards/chosen": -0.013493761420249939, + "rewards/margins": 0.7876949906349182, + "rewards/rejected": -0.8011887669563293, + "step": 741 + }, + { + "epoch": 1.628979143798024, + "grad_norm": 7.245823383331299, + "learning_rate": 5.121980087628802e-07, + "logits/chosen": -2.3491549491882324, + "logits/rejected": -2.3146932125091553, + "logps/chosen": -18.954008102416992, + "logps/rejected": -9.580605506896973, + "loss": 0.7305, + "rewards/accuracies": 0.8125, + "rewards/chosen": 0.012709930539131165, + "rewards/margins": 0.854537844657898, + "rewards/rejected": -0.8418278694152832, + "step": 742 + }, + { + "epoch": 1.6311745334796925, + "grad_norm": 7.404480934143066, + "learning_rate": 5.063326589136211e-07, + "logits/chosen": -2.2543821334838867, + "logits/rejected": -2.2508292198181152, + "logps/chosen": -26.16612434387207, + "logps/rejected": -10.97022533416748, + "loss": 0.6405, + "rewards/accuracies": 0.9375, + "rewards/chosen": 0.09926588833332062, + "rewards/margins": 1.4567351341247559, + "rewards/rejected": -1.3574692010879517, + "step": 743 + }, + { + "epoch": 1.6333699231613612, + "grad_norm": 5.2501654624938965, + "learning_rate": 5.004972994367102e-07, + "logits/chosen": -2.29329252243042, + "logits/rejected": -2.330554723739624, + "logps/chosen": -26.312280654907227, + "logps/rejected": -8.831703186035156, + "loss": 0.6773, + "rewards/accuracies": 0.8125, + "rewards/chosen": 0.10830595344305038, + "rewards/margins": 1.1727694272994995, + "rewards/rejected": -1.0644633769989014, + "step": 744 + }, + { + "epoch": 1.6355653128430296, + "grad_norm": 9.709843635559082, + "learning_rate": 4.946920181123904e-07, + "logits/chosen": -2.3541154861450195, + "logits/rejected": -2.2314836978912354, + "logps/chosen": -21.10788917541504, + "logps/rejected": -11.456113815307617, + "loss": 0.8404, + "rewards/accuracies": 0.84375, + "rewards/chosen": 0.08807536214590073, + "rewards/margins": 0.6532841324806213, + "rewards/rejected": -0.5652087330818176, + "step": 745 + }, + { + "epoch": 1.637760702524698, + "grad_norm": 8.113123893737793, + "learning_rate": 4.889169022684451e-07, + "logits/chosen": -2.32179594039917, + "logits/rejected": -2.2316513061523438, + "logps/chosen": -20.542137145996094, + "logps/rejected": -12.453176498413086, + "loss": 0.8681, + "rewards/accuracies": 0.6875, + "rewards/chosen": 0.016437089070677757, + "rewards/margins": 0.7567267417907715, + "rewards/rejected": -0.7402896881103516, + "step": 746 + }, + { + "epoch": 1.6399560922063667, + "grad_norm": 8.360115051269531, + "learning_rate": 4.831720387788827e-07, + "logits/chosen": -2.3045430183410645, + "logits/rejected": -2.3921265602111816, + "logps/chosen": -27.15880584716797, + "logps/rejected": -11.733960151672363, + "loss": 0.6713, + "rewards/accuracies": 0.875, + "rewards/chosen": 0.12305698543787003, + "rewards/margins": 1.5166301727294922, + "rewards/rejected": -1.39357328414917, + "step": 747 + }, + { + "epoch": 1.6421514818880352, + "grad_norm": 13.261641502380371, + "learning_rate": 4.774575140626317e-07, + "logits/chosen": -2.308157444000244, + "logits/rejected": -2.3163204193115234, + "logps/chosen": -20.00318717956543, + "logps/rejected": -11.708065032958984, + "loss": 0.8056, + "rewards/accuracies": 0.8125, + "rewards/chosen": -0.061866067349910736, + "rewards/margins": 0.865653395652771, + "rewards/rejected": -0.9275195598602295, + "step": 748 + }, + { + "epoch": 1.6443468715697036, + "grad_norm": 6.243008613586426, + "learning_rate": 4.7177341408223996e-07, + "logits/chosen": -2.3002495765686035, + "logits/rejected": -2.263364315032959, + "logps/chosen": -21.880001068115234, + "logps/rejected": -13.372509002685547, + "loss": 0.8127, + "rewards/accuracies": 0.84375, + "rewards/chosen": 0.055353350937366486, + "rewards/margins": 0.9440176486968994, + "rewards/rejected": -0.8886643052101135, + "step": 749 + }, + { + "epoch": 1.6465422612513723, + "grad_norm": 6.65704345703125, + "learning_rate": 4.661198243425813e-07, + "logits/chosen": -2.2934350967407227, + "logits/rejected": -2.315829277038574, + "logps/chosen": -25.759910583496094, + "logps/rejected": -9.104779243469238, + "loss": 0.6088, + "rewards/accuracies": 0.9375, + "rewards/chosen": 0.17460758984088898, + "rewards/margins": 1.2573922872543335, + "rewards/rejected": -1.082784652709961, + "step": 750 + }, + { + "epoch": 1.6487376509330405, + "grad_norm": 7.169310569763184, + "learning_rate": 4.6049682988957037e-07, + "logits/chosen": -2.3058950901031494, + "logits/rejected": -2.335928440093994, + "logps/chosen": -21.361682891845703, + "logps/rejected": -9.361721992492676, + "loss": 0.6872, + "rewards/accuracies": 0.78125, + "rewards/chosen": 0.07690908014774323, + "rewards/margins": 0.9486491680145264, + "rewards/rejected": -0.8717399835586548, + "step": 751 + }, + { + "epoch": 1.6509330406147091, + "grad_norm": 9.812240600585938, + "learning_rate": 4.549045153088813e-07, + "logits/chosen": -2.339125156402588, + "logits/rejected": -2.3041272163391113, + "logps/chosen": -22.344303131103516, + "logps/rejected": -10.660966873168945, + "loss": 0.7153, + "rewards/accuracies": 0.9375, + "rewards/chosen": 0.07993239164352417, + "rewards/margins": 1.1285192966461182, + "rewards/rejected": -1.0485869646072388, + "step": 752 + }, + { + "epoch": 1.6531284302963776, + "grad_norm": 14.382511138916016, + "learning_rate": 4.493429647246775e-07, + "logits/chosen": -2.30961275100708, + "logits/rejected": -2.3265035152435303, + "logps/chosen": -23.634096145629883, + "logps/rejected": -13.76882553100586, + "loss": 0.8332, + "rewards/accuracies": 0.84375, + "rewards/chosen": -0.16528630256652832, + "rewards/margins": 0.9552921056747437, + "rewards/rejected": -1.120578408241272, + "step": 753 + }, + { + "epoch": 1.655323819978046, + "grad_norm": 6.30231237411499, + "learning_rate": 4.438122617983442e-07, + "logits/chosen": -2.3383641242980957, + "logits/rejected": -2.3733930587768555, + "logps/chosen": -22.99635887145996, + "logps/rejected": -9.762690544128418, + "loss": 0.7119, + "rewards/accuracies": 0.6875, + "rewards/chosen": 0.12930777668952942, + "rewards/margins": 1.1605173349380493, + "rewards/rejected": -1.0312095880508423, + "step": 754 + }, + { + "epoch": 1.6575192096597147, + "grad_norm": 9.077330589294434, + "learning_rate": 4.383124897272331e-07, + "logits/chosen": -2.2761359214782715, + "logits/rejected": -2.3204493522644043, + "logps/chosen": -18.578094482421875, + "logps/rejected": -12.708274841308594, + "loss": 0.9148, + "rewards/accuracies": 0.78125, + "rewards/chosen": 0.09242336452007294, + "rewards/margins": 0.571932315826416, + "rewards/rejected": -0.4795089364051819, + "step": 755 + }, + { + "epoch": 1.659714599341383, + "grad_norm": 13.236684799194336, + "learning_rate": 4.3284373124340665e-07, + "logits/chosen": -2.3265042304992676, + "logits/rejected": -2.255683422088623, + "logps/chosen": -21.194488525390625, + "logps/rejected": -12.04061222076416, + "loss": 0.8554, + "rewards/accuracies": 0.78125, + "rewards/chosen": -0.014317229390144348, + "rewards/margins": 0.7051436305046082, + "rewards/rejected": -0.7194608449935913, + "step": 756 + }, + { + "epoch": 1.6619099890230515, + "grad_norm": 10.249890327453613, + "learning_rate": 4.27406068612396e-07, + "logits/chosen": -2.2994461059570312, + "logits/rejected": -2.2328312397003174, + "logps/chosen": -19.790327072143555, + "logps/rejected": -13.203168869018555, + "loss": 0.9126, + "rewards/accuracies": 0.71875, + "rewards/chosen": -0.0464923232793808, + "rewards/margins": 0.7765972018241882, + "rewards/rejected": -0.8230894804000854, + "step": 757 + }, + { + "epoch": 1.6641053787047202, + "grad_norm": 14.354084014892578, + "learning_rate": 4.219995836319632e-07, + "logits/chosen": -2.3120946884155273, + "logits/rejected": -2.333422899246216, + "logps/chosen": -19.07431411743164, + "logps/rejected": -8.69210147857666, + "loss": 0.7516, + "rewards/accuracies": 0.75, + "rewards/chosen": 0.051367685198783875, + "rewards/margins": 0.9122320413589478, + "rewards/rejected": -0.8608644008636475, + "step": 758 + }, + { + "epoch": 1.6663007683863886, + "grad_norm": 8.517876625061035, + "learning_rate": 4.166243576308712e-07, + "logits/chosen": -2.274487257003784, + "logits/rejected": -2.2832393646240234, + "logps/chosen": -21.78612518310547, + "logps/rejected": -9.265876770019531, + "loss": 0.6727, + "rewards/accuracies": 0.875, + "rewards/chosen": 0.06986172497272491, + "rewards/margins": 1.0944383144378662, + "rewards/rejected": -1.0245764255523682, + "step": 759 + }, + { + "epoch": 1.668496158068057, + "grad_norm": 12.419169425964355, + "learning_rate": 4.1128047146765936e-07, + "logits/chosen": -2.4093031883239746, + "logits/rejected": -2.335050344467163, + "logps/chosen": -19.587875366210938, + "logps/rejected": -10.676223754882812, + "loss": 0.8096, + "rewards/accuracies": 0.71875, + "rewards/chosen": -0.04506263881921768, + "rewards/margins": 0.6117575168609619, + "rewards/rejected": -0.6568201780319214, + "step": 760 + }, + { + "epoch": 1.6706915477497257, + "grad_norm": 8.814732551574707, + "learning_rate": 4.0596800552942655e-07, + "logits/chosen": -2.324427366256714, + "logits/rejected": -2.3086276054382324, + "logps/chosen": -25.435884475708008, + "logps/rejected": -10.919512748718262, + "loss": 0.7239, + "rewards/accuracies": 0.84375, + "rewards/chosen": 0.10456670075654984, + "rewards/margins": 1.172752857208252, + "rewards/rejected": -1.0681862831115723, + "step": 761 + }, + { + "epoch": 1.672886937431394, + "grad_norm": 6.977013111114502, + "learning_rate": 4.006870397306256e-07, + "logits/chosen": -2.3290557861328125, + "logits/rejected": -2.408146381378174, + "logps/chosen": -23.841257095336914, + "logps/rejected": -10.088373184204102, + "loss": 0.686, + "rewards/accuracies": 0.8125, + "rewards/chosen": 0.08101833611726761, + "rewards/margins": 1.169783115386963, + "rewards/rejected": -1.0887647867202759, + "step": 762 + }, + { + "epoch": 1.6750823271130626, + "grad_norm": 8.580155372619629, + "learning_rate": 3.954376535118562e-07, + "logits/chosen": -2.365082025527954, + "logits/rejected": -2.3728320598602295, + "logps/chosen": -23.600791931152344, + "logps/rejected": -12.741727828979492, + "loss": 0.7832, + "rewards/accuracies": 0.78125, + "rewards/chosen": -0.04538576677441597, + "rewards/margins": 0.9993623495101929, + "rewards/rejected": -1.044748067855835, + "step": 763 + }, + { + "epoch": 1.677277716794731, + "grad_norm": 7.796408653259277, + "learning_rate": 3.9021992583867326e-07, + "logits/chosen": -2.3659496307373047, + "logits/rejected": -2.3780646324157715, + "logps/chosen": -28.731002807617188, + "logps/rejected": -19.395036697387695, + "loss": 0.9424, + "rewards/accuracies": 0.8125, + "rewards/chosen": -0.0875871330499649, + "rewards/margins": 1.2551369667053223, + "rewards/rejected": -1.3427242040634155, + "step": 764 + }, + { + "epoch": 1.6794731064763995, + "grad_norm": 7.225224018096924, + "learning_rate": 3.8503393520039734e-07, + "logits/chosen": -2.2901337146759033, + "logits/rejected": -2.236420154571533, + "logps/chosen": -26.338916778564453, + "logps/rejected": -12.186105728149414, + "loss": 0.7196, + "rewards/accuracies": 0.8125, + "rewards/chosen": -0.03397097438573837, + "rewards/margins": 1.1694746017456055, + "rewards/rejected": -1.2034454345703125, + "step": 765 + }, + { + "epoch": 1.6816684961580681, + "grad_norm": 6.434988021850586, + "learning_rate": 3.798797596089351e-07, + "logits/chosen": -2.2957446575164795, + "logits/rejected": -2.3498806953430176, + "logps/chosen": -21.051616668701172, + "logps/rejected": -10.8973388671875, + "loss": 0.7672, + "rewards/accuracies": 0.75, + "rewards/chosen": 0.10705845057964325, + "rewards/margins": 0.9983046054840088, + "rewards/rejected": -0.8912461996078491, + "step": 766 + }, + { + "epoch": 1.6838638858397366, + "grad_norm": 6.382670879364014, + "learning_rate": 3.7475747659760504e-07, + "logits/chosen": -2.2628824710845947, + "logits/rejected": -2.3416075706481934, + "logps/chosen": -22.019775390625, + "logps/rejected": -9.292243957519531, + "loss": 0.7142, + "rewards/accuracies": 0.78125, + "rewards/chosen": 0.11236327141523361, + "rewards/margins": 1.0206373929977417, + "rewards/rejected": -0.9082741737365723, + "step": 767 + }, + { + "epoch": 1.686059275521405, + "grad_norm": 8.05347728729248, + "learning_rate": 3.6966716321997195e-07, + "logits/chosen": -2.271007537841797, + "logits/rejected": -2.3000264167785645, + "logps/chosen": -23.111034393310547, + "logps/rejected": -13.595745086669922, + "loss": 0.8505, + "rewards/accuracies": 0.875, + "rewards/chosen": -0.06358395516872406, + "rewards/margins": 0.9935811758041382, + "rewards/rejected": -1.0571651458740234, + "step": 768 + }, + { + "epoch": 1.6882546652030737, + "grad_norm": 5.555905818939209, + "learning_rate": 3.646088960486863e-07, + "logits/chosen": -2.3269221782684326, + "logits/rejected": -2.311516761779785, + "logps/chosen": -23.083637237548828, + "logps/rejected": -9.06369686126709, + "loss": 0.6524, + "rewards/accuracies": 0.875, + "rewards/chosen": 0.11144015938043594, + "rewards/margins": 1.1470022201538086, + "rewards/rejected": -1.0355621576309204, + "step": 769 + }, + { + "epoch": 1.690450054884742, + "grad_norm": 5.618292331695557, + "learning_rate": 3.595827511743341e-07, + "logits/chosen": -2.3102753162384033, + "logits/rejected": -2.3263516426086426, + "logps/chosen": -20.93044090270996, + "logps/rejected": -8.075803756713867, + "loss": 0.6733, + "rewards/accuracies": 0.84375, + "rewards/chosen": 0.22235707938671112, + "rewards/margins": 0.9959395527839661, + "rewards/rejected": -0.7735824584960938, + "step": 770 + }, + { + "epoch": 1.6926454445664105, + "grad_norm": 6.7625579833984375, + "learning_rate": 3.545888042042914e-07, + "logits/chosen": -2.2616472244262695, + "logits/rejected": -2.22017765045166, + "logps/chosen": -22.324371337890625, + "logps/rejected": -11.00965690612793, + "loss": 0.6918, + "rewards/accuracies": 0.90625, + "rewards/chosen": 0.14851540327072144, + "rewards/margins": 1.0847861766815186, + "rewards/rejected": -0.9362708330154419, + "step": 771 + }, + { + "epoch": 1.6948408342480792, + "grad_norm": 7.646176338195801, + "learning_rate": 3.4962713026158697e-07, + "logits/chosen": -2.302961826324463, + "logits/rejected": -2.2982404232025146, + "logps/chosen": -29.35641098022461, + "logps/rejected": -12.110550880432129, + "loss": 0.6514, + "rewards/accuracies": 0.90625, + "rewards/chosen": 0.09808993339538574, + "rewards/margins": 1.7121081352233887, + "rewards/rejected": -1.6140180826187134, + "step": 772 + }, + { + "epoch": 1.6970362239297474, + "grad_norm": 7.379415512084961, + "learning_rate": 3.446978039837726e-07, + "logits/chosen": -2.359391689300537, + "logits/rejected": -2.312047004699707, + "logps/chosen": -30.472061157226562, + "logps/rejected": -12.87568473815918, + "loss": 0.741, + "rewards/accuracies": 0.6875, + "rewards/chosen": -0.04519427567720413, + "rewards/margins": 1.093401551246643, + "rewards/rejected": -1.1385958194732666, + "step": 773 + }, + { + "epoch": 1.699231613611416, + "grad_norm": 12.749286651611328, + "learning_rate": 3.3980089952179886e-07, + "logits/chosen": -2.2937393188476562, + "logits/rejected": -2.345874071121216, + "logps/chosen": -22.466127395629883, + "logps/rejected": -9.391763687133789, + "loss": 0.7854, + "rewards/accuracies": 0.75, + "rewards/chosen": 0.05167246237397194, + "rewards/margins": 0.8294394016265869, + "rewards/rejected": -0.7777669429779053, + "step": 774 + }, + { + "epoch": 1.7014270032930845, + "grad_norm": 11.830827713012695, + "learning_rate": 3.3493649053890325e-07, + "logits/chosen": -2.3040733337402344, + "logits/rejected": -2.2944912910461426, + "logps/chosen": -21.27988052368164, + "logps/rejected": -10.81669807434082, + "loss": 0.8225, + "rewards/accuracies": 0.8125, + "rewards/chosen": -0.027055367827415466, + "rewards/margins": 0.7839744687080383, + "rewards/rejected": -0.811029851436615, + "step": 775 + }, + { + "epoch": 1.703622392974753, + "grad_norm": 6.718094348907471, + "learning_rate": 3.3010465020949815e-07, + "logits/chosen": -2.2920472621917725, + "logits/rejected": -2.3191323280334473, + "logps/chosen": -22.437255859375, + "logps/rejected": -9.567081451416016, + "loss": 0.6813, + "rewards/accuracies": 0.84375, + "rewards/chosen": 0.11176694929599762, + "rewards/margins": 1.091386318206787, + "rewards/rejected": -0.9796194434165955, + "step": 776 + }, + { + "epoch": 1.7058177826564216, + "grad_norm": 8.094005584716797, + "learning_rate": 3.2530545121807145e-07, + "logits/chosen": -2.2624449729919434, + "logits/rejected": -2.3998560905456543, + "logps/chosen": -23.40216636657715, + "logps/rejected": -14.835620880126953, + "loss": 0.8838, + "rewards/accuracies": 0.71875, + "rewards/chosen": 0.24612218141555786, + "rewards/margins": 1.135087251663208, + "rewards/rejected": -0.8889651298522949, + "step": 777 + }, + { + "epoch": 1.70801317233809, + "grad_norm": 6.472238063812256, + "learning_rate": 3.2053896575809426e-07, + "logits/chosen": -2.340385913848877, + "logits/rejected": -2.26383638381958, + "logps/chosen": -18.576879501342773, + "logps/rejected": -9.460345268249512, + "loss": 0.7067, + "rewards/accuracies": 0.875, + "rewards/chosen": 0.12470845878124237, + "rewards/margins": 0.7781118750572205, + "rewards/rejected": -0.6534034013748169, + "step": 778 + }, + { + "epoch": 1.7102085620197585, + "grad_norm": 12.53104019165039, + "learning_rate": 3.158052655309332e-07, + "logits/chosen": -2.357607841491699, + "logits/rejected": -2.3493804931640625, + "logps/chosen": -21.966833114624023, + "logps/rejected": -11.84433650970459, + "loss": 0.736, + "rewards/accuracies": 0.84375, + "rewards/chosen": 0.04521423578262329, + "rewards/margins": 1.1221758127212524, + "rewards/rejected": -1.0769615173339844, + "step": 779 + }, + { + "epoch": 1.7124039517014271, + "grad_norm": 7.429686546325684, + "learning_rate": 3.111044217447731e-07, + "logits/chosen": -2.3728270530700684, + "logits/rejected": -2.277714490890503, + "logps/chosen": -22.915681838989258, + "logps/rejected": -13.93270206451416, + "loss": 0.9084, + "rewards/accuracies": 0.78125, + "rewards/chosen": 0.03567001596093178, + "rewards/margins": 0.7132501006126404, + "rewards/rejected": -0.6775801181793213, + "step": 780 + }, + { + "epoch": 1.7145993413830956, + "grad_norm": 7.569916248321533, + "learning_rate": 3.0643650511354483e-07, + "logits/chosen": -2.32142972946167, + "logits/rejected": -2.246826410293579, + "logps/chosen": -21.596290588378906, + "logps/rejected": -14.724076271057129, + "loss": 0.8821, + "rewards/accuracies": 0.6875, + "rewards/chosen": 0.007006680592894554, + "rewards/margins": 0.6938904523849487, + "rewards/rejected": -0.6868837475776672, + "step": 781 + }, + { + "epoch": 1.716794731064764, + "grad_norm": 10.38869857788086, + "learning_rate": 3.01801585855864e-07, + "logits/chosen": -2.2968714237213135, + "logits/rejected": -2.301407814025879, + "logps/chosen": -22.191951751708984, + "logps/rejected": -13.601604461669922, + "loss": 0.8494, + "rewards/accuracies": 0.8125, + "rewards/chosen": 0.008423591032624245, + "rewards/margins": 0.976859986782074, + "rewards/rejected": -0.9684363603591919, + "step": 782 + }, + { + "epoch": 1.7189901207464326, + "grad_norm": 6.449059963226318, + "learning_rate": 2.9719973369397025e-07, + "logits/chosen": -2.306488513946533, + "logits/rejected": -2.240659713745117, + "logps/chosen": -22.008119583129883, + "logps/rejected": -15.796365737915039, + "loss": 1.0061, + "rewards/accuracies": 0.53125, + "rewards/chosen": -0.04236738011240959, + "rewards/margins": 0.4466336965560913, + "rewards/rejected": -0.4890010952949524, + "step": 783 + }, + { + "epoch": 1.7211855104281009, + "grad_norm": 9.239655494689941, + "learning_rate": 2.9263101785268253e-07, + "logits/chosen": -2.325709342956543, + "logits/rejected": -2.313990592956543, + "logps/chosen": -22.03015899658203, + "logps/rejected": -13.100378036499023, + "loss": 0.8501, + "rewards/accuracies": 0.78125, + "rewards/chosen": 0.026344943791627884, + "rewards/margins": 0.8191928267478943, + "rewards/rejected": -0.7928478717803955, + "step": 784 + }, + { + "epoch": 1.7233809001097695, + "grad_norm": 26.027679443359375, + "learning_rate": 2.880955070583555e-07, + "logits/chosen": -2.3623580932617188, + "logits/rejected": -2.328763008117676, + "logps/chosen": -20.698711395263672, + "logps/rejected": -17.81403350830078, + "loss": 1.0792, + "rewards/accuracies": 0.625, + "rewards/chosen": -0.38904720544815063, + "rewards/margins": 0.5637388825416565, + "rewards/rejected": -0.9527859687805176, + "step": 785 + }, + { + "epoch": 1.725576289791438, + "grad_norm": 6.592658519744873, + "learning_rate": 2.835932695378474e-07, + "logits/chosen": -2.3217051029205322, + "logits/rejected": -2.2342302799224854, + "logps/chosen": -29.108089447021484, + "logps/rejected": -12.419905662536621, + "loss": 0.6403, + "rewards/accuracies": 0.90625, + "rewards/chosen": 0.17239367961883545, + "rewards/margins": 1.5372717380523682, + "rewards/rejected": -1.3648779392242432, + "step": 786 + }, + { + "epoch": 1.7277716794731064, + "grad_norm": 7.226133346557617, + "learning_rate": 2.791243730174903e-07, + "logits/chosen": -2.225740909576416, + "logits/rejected": -2.3681936264038086, + "logps/chosen": -26.986141204833984, + "logps/rejected": -13.851487159729004, + "loss": 0.7736, + "rewards/accuracies": 0.875, + "rewards/chosen": 0.07572560012340546, + "rewards/margins": 1.4555649757385254, + "rewards/rejected": -1.3798394203186035, + "step": 787 + }, + { + "epoch": 1.729967069154775, + "grad_norm": 9.024186134338379, + "learning_rate": 2.7468888472207606e-07, + "logits/chosen": -2.252058267593384, + "logits/rejected": -2.3360979557037354, + "logps/chosen": -26.153547286987305, + "logps/rejected": -15.990081787109375, + "loss": 0.7992, + "rewards/accuracies": 0.875, + "rewards/chosen": 0.12151572108268738, + "rewards/margins": 1.3226542472839355, + "rewards/rejected": -1.2011386156082153, + "step": 788 + }, + { + "epoch": 1.7321624588364435, + "grad_norm": 9.212321281433105, + "learning_rate": 2.702868713738427e-07, + "logits/chosen": -2.3098626136779785, + "logits/rejected": -2.285076379776001, + "logps/chosen": -22.022165298461914, + "logps/rejected": -12.408515930175781, + "loss": 0.7779, + "rewards/accuracies": 0.875, + "rewards/chosen": 0.01777059957385063, + "rewards/margins": 0.8607884049415588, + "rewards/rejected": -0.8430178165435791, + "step": 789 + }, + { + "epoch": 1.734357848518112, + "grad_norm": 11.648327827453613, + "learning_rate": 2.6591839919146963e-07, + "logits/chosen": -2.3739309310913086, + "logits/rejected": -2.297510862350464, + "logps/chosen": -20.9400691986084, + "logps/rejected": -13.20623779296875, + "loss": 0.8605, + "rewards/accuracies": 0.6875, + "rewards/chosen": -0.04880945757031441, + "rewards/margins": 0.8373270630836487, + "rewards/rejected": -0.886136531829834, + "step": 790 + }, + { + "epoch": 1.7365532381997806, + "grad_norm": 7.143850803375244, + "learning_rate": 2.6158353388908294e-07, + "logits/chosen": -2.2994394302368164, + "logits/rejected": -2.29927396774292, + "logps/chosen": -23.870580673217773, + "logps/rejected": -13.661394119262695, + "loss": 0.8087, + "rewards/accuracies": 0.84375, + "rewards/chosen": 0.06044537574052811, + "rewards/margins": 1.079881191253662, + "rewards/rejected": -1.0194358825683594, + "step": 791 + }, + { + "epoch": 1.7387486278814488, + "grad_norm": 10.934171676635742, + "learning_rate": 2.5728234067526695e-07, + "logits/chosen": -2.2999086380004883, + "logits/rejected": -2.3465735912323, + "logps/chosen": -27.578174591064453, + "logps/rejected": -10.039525985717773, + "loss": 0.7328, + "rewards/accuracies": 0.84375, + "rewards/chosen": 0.04027145728468895, + "rewards/margins": 1.0723406076431274, + "rewards/rejected": -1.032069206237793, + "step": 792 + }, + { + "epoch": 1.7409440175631175, + "grad_norm": 6.294828414916992, + "learning_rate": 2.53014884252083e-07, + "logits/chosen": -2.320455312728882, + "logits/rejected": -2.2217442989349365, + "logps/chosen": -24.052602767944336, + "logps/rejected": -11.206563949584961, + "loss": 0.7004, + "rewards/accuracies": 0.84375, + "rewards/chosen": 0.1112891435623169, + "rewards/margins": 0.9485012292861938, + "rewards/rejected": -0.837212085723877, + "step": 793 + }, + { + "epoch": 1.743139407244786, + "grad_norm": 16.46431541442871, + "learning_rate": 2.487812288140945e-07, + "logits/chosen": -2.305223226547241, + "logits/rejected": -2.2811007499694824, + "logps/chosen": -20.011690139770508, + "logps/rejected": -10.267993927001953, + "loss": 0.7854, + "rewards/accuracies": 0.71875, + "rewards/chosen": 0.06191223859786987, + "rewards/margins": 0.9365310668945312, + "rewards/rejected": -0.8746187686920166, + "step": 794 + }, + { + "epoch": 1.7453347969264543, + "grad_norm": 10.158160209655762, + "learning_rate": 2.445814380474057e-07, + "logits/chosen": -2.2518134117126465, + "logits/rejected": -2.3549113273620605, + "logps/chosen": -22.6698055267334, + "logps/rejected": -11.738947868347168, + "loss": 0.77, + "rewards/accuracies": 0.875, + "rewards/chosen": 0.016972137615084648, + "rewards/margins": 1.0493351221084595, + "rewards/rejected": -1.032362937927246, + "step": 795 + }, + { + "epoch": 1.747530186608123, + "grad_norm": 17.583005905151367, + "learning_rate": 2.404155751286988e-07, + "logits/chosen": -2.2456798553466797, + "logits/rejected": -2.2801907062530518, + "logps/chosen": -22.485057830810547, + "logps/rejected": -17.982084274291992, + "loss": 1.1674, + "rewards/accuracies": 0.6875, + "rewards/chosen": -0.14132660627365112, + "rewards/margins": 0.32514142990112305, + "rewards/rejected": -0.46646803617477417, + "step": 796 + }, + { + "epoch": 1.7497255762897914, + "grad_norm": 7.677198886871338, + "learning_rate": 2.3628370272428562e-07, + "logits/chosen": -2.2841689586639404, + "logits/rejected": -2.302462577819824, + "logps/chosen": -22.194114685058594, + "logps/rejected": -11.15237045288086, + "loss": 0.716, + "rewards/accuracies": 0.84375, + "rewards/chosen": 0.15826064348220825, + "rewards/margins": 0.9130882024765015, + "rewards/rejected": -0.7548274993896484, + "step": 797 + }, + { + "epoch": 1.7519209659714599, + "grad_norm": 12.027734756469727, + "learning_rate": 2.3218588298916545e-07, + "logits/chosen": -2.292127847671509, + "logits/rejected": -2.3266634941101074, + "logps/chosen": -26.069114685058594, + "logps/rejected": -14.663209915161133, + "loss": 0.8585, + "rewards/accuracies": 0.8125, + "rewards/chosen": 0.13133227825164795, + "rewards/margins": 0.8031163811683655, + "rewards/rejected": -0.6717841029167175, + "step": 798 + }, + { + "epoch": 1.7541163556531285, + "grad_norm": 9.412739753723145, + "learning_rate": 2.281221775660894e-07, + "logits/chosen": -2.3051884174346924, + "logits/rejected": -2.3385562896728516, + "logps/chosen": -18.961666107177734, + "logps/rejected": -12.03260612487793, + "loss": 0.8988, + "rewards/accuracies": 0.8125, + "rewards/chosen": -0.15662559866905212, + "rewards/margins": 0.640987753868103, + "rewards/rejected": -0.7976133227348328, + "step": 799 + }, + { + "epoch": 1.756311745334797, + "grad_norm": 7.2472662925720215, + "learning_rate": 2.240926475846336e-07, + "logits/chosen": -2.3602917194366455, + "logits/rejected": -2.2820887565612793, + "logps/chosen": -27.15008544921875, + "logps/rejected": -16.696186065673828, + "loss": 0.8362, + "rewards/accuracies": 0.8125, + "rewards/chosen": 0.03183577209711075, + "rewards/margins": 1.2374306917190552, + "rewards/rejected": -1.2055950164794922, + "step": 800 + }, + { + "epoch": 1.7585071350164654, + "grad_norm": 9.894173622131348, + "learning_rate": 2.2009735366027796e-07, + "logits/chosen": -2.288947820663452, + "logits/rejected": -2.3076653480529785, + "logps/chosen": -34.05409240722656, + "logps/rejected": -11.510194778442383, + "loss": 0.5484, + "rewards/accuracies": 0.90625, + "rewards/chosen": 0.10917800664901733, + "rewards/margins": 2.0579628944396973, + "rewards/rejected": -1.948784589767456, + "step": 801 + }, + { + "epoch": 1.760702524698134, + "grad_norm": 6.294257164001465, + "learning_rate": 2.1613635589349756e-07, + "logits/chosen": -2.2891159057617188, + "logits/rejected": -2.408547878265381, + "logps/chosen": -28.716947555541992, + "logps/rejected": -13.035652160644531, + "loss": 0.7472, + "rewards/accuracies": 0.875, + "rewards/chosen": 0.02225133776664734, + "rewards/margins": 1.5577504634857178, + "rewards/rejected": -1.5354992151260376, + "step": 802 + }, + { + "epoch": 1.7628979143798023, + "grad_norm": 6.781076431274414, + "learning_rate": 2.1220971386885557e-07, + "logits/chosen": -2.2559494972229004, + "logits/rejected": -2.339909315109253, + "logps/chosen": -28.129169464111328, + "logps/rejected": -13.074584007263184, + "loss": 0.7831, + "rewards/accuracies": 0.78125, + "rewards/chosen": 0.0403079055249691, + "rewards/margins": 1.390425205230713, + "rewards/rejected": -1.3501174449920654, + "step": 803 + }, + { + "epoch": 1.765093304061471, + "grad_norm": 8.248476028442383, + "learning_rate": 2.0831748665410768e-07, + "logits/chosen": -2.2961597442626953, + "logits/rejected": -2.348660945892334, + "logps/chosen": -24.3775634765625, + "logps/rejected": -12.936763763427734, + "loss": 0.7833, + "rewards/accuracies": 0.78125, + "rewards/chosen": -0.06602227687835693, + "rewards/margins": 1.0288200378417969, + "rewards/rejected": -1.0948423147201538, + "step": 804 + }, + { + "epoch": 1.7672886937431396, + "grad_norm": 10.043867111206055, + "learning_rate": 2.044597327993153e-07, + "logits/chosen": -2.2116641998291016, + "logits/rejected": -2.293318033218384, + "logps/chosen": -24.019229888916016, + "logps/rejected": -12.467460632324219, + "loss": 0.8294, + "rewards/accuracies": 0.78125, + "rewards/chosen": 0.06100135296583176, + "rewards/margins": 0.9037472605705261, + "rewards/rejected": -0.8427459001541138, + "step": 805 + }, + { + "epoch": 1.7694840834248078, + "grad_norm": 6.782207012176514, + "learning_rate": 2.0063651033596143e-07, + "logits/chosen": -2.248141050338745, + "logits/rejected": -2.336304187774658, + "logps/chosen": -27.149402618408203, + "logps/rejected": -11.001330375671387, + "loss": 0.6133, + "rewards/accuracies": 0.90625, + "rewards/chosen": 0.06859119236469269, + "rewards/margins": 1.479909062385559, + "rewards/rejected": -1.4113179445266724, + "step": 806 + }, + { + "epoch": 1.7716794731064764, + "grad_norm": 8.469517707824707, + "learning_rate": 1.968478767760812e-07, + "logits/chosen": -2.323946475982666, + "logits/rejected": -2.290382146835327, + "logps/chosen": -23.23932647705078, + "logps/rejected": -9.99721908569336, + "loss": 0.6675, + "rewards/accuracies": 0.875, + "rewards/chosen": 0.06563490629196167, + "rewards/margins": 1.2968019247055054, + "rewards/rejected": -1.2311670780181885, + "step": 807 + }, + { + "epoch": 1.7738748627881449, + "grad_norm": 6.731967449188232, + "learning_rate": 1.9309388911139427e-07, + "logits/chosen": -2.2910850048065186, + "logits/rejected": -2.3337833881378174, + "logps/chosen": -24.140378952026367, + "logps/rejected": -8.914945602416992, + "loss": 0.6769, + "rewards/accuracies": 0.8125, + "rewards/chosen": 0.12325699627399445, + "rewards/margins": 1.3196156024932861, + "rewards/rejected": -1.1963586807250977, + "step": 808 + }, + { + "epoch": 1.7760702524698133, + "grad_norm": 14.611775398254395, + "learning_rate": 1.893746038124497e-07, + "logits/chosen": -2.3058838844299316, + "logits/rejected": -2.3319802284240723, + "logps/chosen": -18.569305419921875, + "logps/rejected": -12.480390548706055, + "loss": 0.8895, + "rewards/accuracies": 0.71875, + "rewards/chosen": -0.05126887932419777, + "rewards/margins": 0.6904062032699585, + "rewards/rejected": -0.7416751384735107, + "step": 809 + }, + { + "epoch": 1.778265642151482, + "grad_norm": 9.839641571044922, + "learning_rate": 1.8569007682777417e-07, + "logits/chosen": -2.3887779712677, + "logits/rejected": -2.354360818862915, + "logps/chosen": -22.196651458740234, + "logps/rejected": -11.59244155883789, + "loss": 0.7609, + "rewards/accuracies": 0.90625, + "rewards/chosen": 0.058909520506858826, + "rewards/margins": 1.0041024684906006, + "rewards/rejected": -0.9451929330825806, + "step": 810 + }, + { + "epoch": 1.7804610318331504, + "grad_norm": 8.44560718536377, + "learning_rate": 1.8204036358303173e-07, + "logits/chosen": -2.254578113555908, + "logits/rejected": -2.321296453475952, + "logps/chosen": -24.86836814880371, + "logps/rejected": -10.461334228515625, + "loss": 0.6586, + "rewards/accuracies": 0.875, + "rewards/chosen": 0.007388472557067871, + "rewards/margins": 1.4799784421920776, + "rewards/rejected": -1.4725899696350098, + "step": 811 + }, + { + "epoch": 1.7826564215148188, + "grad_norm": 7.611237049102783, + "learning_rate": 1.784255189801895e-07, + "logits/chosen": -2.364264726638794, + "logits/rejected": -2.336348056793213, + "logps/chosen": -22.660598754882812, + "logps/rejected": -12.24748420715332, + "loss": 0.8079, + "rewards/accuracies": 0.75, + "rewards/chosen": 0.006627365946769714, + "rewards/margins": 1.0588319301605225, + "rewards/rejected": -1.0522046089172363, + "step": 812 + }, + { + "epoch": 1.7848518111964875, + "grad_norm": 12.086380958557129, + "learning_rate": 1.7484559739669226e-07, + "logits/chosen": -2.4495887756347656, + "logits/rejected": -2.2654292583465576, + "logps/chosen": -17.38649559020996, + "logps/rejected": -12.829127311706543, + "loss": 0.9107, + "rewards/accuracies": 0.6875, + "rewards/chosen": -0.031765177845954895, + "rewards/margins": 0.523868978023529, + "rewards/rejected": -0.5556341409683228, + "step": 813 + }, + { + "epoch": 1.7870472008781557, + "grad_norm": 12.084946632385254, + "learning_rate": 1.713006526846439e-07, + "logits/chosen": -2.2484347820281982, + "logits/rejected": -2.321169376373291, + "logps/chosen": -25.002975463867188, + "logps/rejected": -12.445917129516602, + "loss": 0.7518, + "rewards/accuracies": 0.84375, + "rewards/chosen": 0.024622926488518715, + "rewards/margins": 1.2067277431488037, + "rewards/rejected": -1.1821048259735107, + "step": 814 + }, + { + "epoch": 1.7892425905598244, + "grad_norm": 4.94568395614624, + "learning_rate": 1.6779073816999864e-07, + "logits/chosen": -2.3164916038513184, + "logits/rejected": -2.2931370735168457, + "logps/chosen": -17.374296188354492, + "logps/rejected": -9.248968124389648, + "loss": 0.8257, + "rewards/accuracies": 0.8125, + "rewards/chosen": 0.08453521132469177, + "rewards/margins": 0.6115543842315674, + "rewards/rejected": -0.5270192623138428, + "step": 815 + }, + { + "epoch": 1.7914379802414928, + "grad_norm": 9.331941604614258, + "learning_rate": 1.6431590665175663e-07, + "logits/chosen": -2.3295576572418213, + "logits/rejected": -2.291257858276367, + "logps/chosen": -21.367877960205078, + "logps/rejected": -10.068951606750488, + "loss": 0.8291, + "rewards/accuracies": 0.78125, + "rewards/chosen": -0.035824254155159, + "rewards/margins": 0.9504915475845337, + "rewards/rejected": -0.9863157868385315, + "step": 816 + }, + { + "epoch": 1.7936333699231612, + "grad_norm": 8.115439414978027, + "learning_rate": 1.6087621040117157e-07, + "logits/chosen": -2.4213380813598633, + "logits/rejected": -2.3565773963928223, + "logps/chosen": -23.354394912719727, + "logps/rejected": -11.980990409851074, + "loss": 0.7478, + "rewards/accuracies": 0.78125, + "rewards/chosen": 0.1042669266462326, + "rewards/margins": 1.081133484840393, + "rewards/rejected": -0.9768665432929993, + "step": 817 + }, + { + "epoch": 1.79582875960483, + "grad_norm": 13.605571746826172, + "learning_rate": 1.5747170116096333e-07, + "logits/chosen": -2.322146415710449, + "logits/rejected": -2.332505226135254, + "logps/chosen": -18.397136688232422, + "logps/rejected": -9.858526229858398, + "loss": 0.8103, + "rewards/accuracies": 0.65625, + "rewards/chosen": -0.020177535712718964, + "rewards/margins": 0.7551686763763428, + "rewards/rejected": -0.7753462195396423, + "step": 818 + }, + { + "epoch": 1.7980241492864983, + "grad_norm": 10.637507438659668, + "learning_rate": 1.541024301445404e-07, + "logits/chosen": -2.309520959854126, + "logits/rejected": -2.3439934253692627, + "logps/chosen": -23.78762435913086, + "logps/rejected": -16.25725746154785, + "loss": 0.888, + "rewards/accuracies": 0.875, + "rewards/chosen": 0.0241999551653862, + "rewards/margins": 0.9779789447784424, + "rewards/rejected": -0.9537790417671204, + "step": 819 + }, + { + "epoch": 1.8002195389681668, + "grad_norm": 5.118252754211426, + "learning_rate": 1.507684480352292e-07, + "logits/chosen": -2.3868606090545654, + "logits/rejected": -2.3554153442382812, + "logps/chosen": -31.286975860595703, + "logps/rejected": -9.514917373657227, + "loss": 0.5666, + "rewards/accuracies": 0.90625, + "rewards/chosen": 0.15696834027767181, + "rewards/margins": 1.8301570415496826, + "rewards/rejected": -1.6731886863708496, + "step": 820 + }, + { + "epoch": 1.8024149286498354, + "grad_norm": 9.137975692749023, + "learning_rate": 1.4746980498551112e-07, + "logits/chosen": -2.2558650970458984, + "logits/rejected": -2.318466901779175, + "logps/chosen": -25.57559585571289, + "logps/rejected": -14.859004974365234, + "loss": 0.8932, + "rewards/accuracies": 0.78125, + "rewards/chosen": 0.11497283726930618, + "rewards/margins": 1.0942399501800537, + "rewards/rejected": -0.9792671203613281, + "step": 821 + }, + { + "epoch": 1.8046103183315039, + "grad_norm": 8.669699668884277, + "learning_rate": 1.4420655061626933e-07, + "logits/chosen": -2.3203203678131104, + "logits/rejected": -2.3640811443328857, + "logps/chosen": -21.141521453857422, + "logps/rejected": -10.256974220275879, + "loss": 0.7005, + "rewards/accuracies": 0.8125, + "rewards/chosen": -0.020266126841306686, + "rewards/margins": 0.8457608819007874, + "rewards/rejected": -0.8660269975662231, + "step": 822 + }, + { + "epoch": 1.8068057080131723, + "grad_norm": 8.92252254486084, + "learning_rate": 1.4097873401604124e-07, + "logits/chosen": -2.3813881874084473, + "logits/rejected": -2.2835211753845215, + "logps/chosen": -19.4898681640625, + "logps/rejected": -9.444177627563477, + "loss": 0.7318, + "rewards/accuracies": 0.8125, + "rewards/chosen": 0.0626252144575119, + "rewards/margins": 0.882063090801239, + "rewards/rejected": -0.8194378018379211, + "step": 823 + }, + { + "epoch": 1.809001097694841, + "grad_norm": 11.689897537231445, + "learning_rate": 1.3778640374027984e-07, + "logits/chosen": -2.2980411052703857, + "logits/rejected": -2.31156587600708, + "logps/chosen": -21.046966552734375, + "logps/rejected": -11.255702018737793, + "loss": 0.811, + "rewards/accuracies": 0.8125, + "rewards/chosen": -0.02777668461203575, + "rewards/margins": 0.8196637630462646, + "rewards/rejected": -0.8474404811859131, + "step": 824 + }, + { + "epoch": 1.8111964873765092, + "grad_norm": 12.901611328125, + "learning_rate": 1.3462960781062433e-07, + "logits/chosen": -2.2532966136932373, + "logits/rejected": -2.373487949371338, + "logps/chosen": -22.69452667236328, + "logps/rejected": -11.57452392578125, + "loss": 0.753, + "rewards/accuracies": 0.8125, + "rewards/chosen": 0.10819564014673233, + "rewards/margins": 1.194549798965454, + "rewards/rejected": -1.0863540172576904, + "step": 825 + }, + { + "epoch": 1.8133918770581778, + "grad_norm": 9.201147079467773, + "learning_rate": 1.31508393714177e-07, + "logits/chosen": -2.275195837020874, + "logits/rejected": -2.2960829734802246, + "logps/chosen": -17.56732749938965, + "logps/rejected": -12.735163688659668, + "loss": 0.9003, + "rewards/accuracies": 0.71875, + "rewards/chosen": 0.07135310769081116, + "rewards/margins": 0.8148505687713623, + "rewards/rejected": -0.7434974908828735, + "step": 826 + }, + { + "epoch": 1.8155872667398463, + "grad_norm": 9.020854949951172, + "learning_rate": 1.2842280840278998e-07, + "logits/chosen": -2.3004255294799805, + "logits/rejected": -2.342841386795044, + "logps/chosen": -29.2739200592041, + "logps/rejected": -11.778656005859375, + "loss": 0.6985, + "rewards/accuracies": 0.90625, + "rewards/chosen": -0.0008747577667236328, + "rewards/margins": 1.4489872455596924, + "rewards/rejected": -1.4498621225357056, + "step": 827 + }, + { + "epoch": 1.8177826564215147, + "grad_norm": 10.636558532714844, + "learning_rate": 1.2537289829235633e-07, + "logits/chosen": -2.4058544635772705, + "logits/rejected": -2.349201202392578, + "logps/chosen": -24.120586395263672, + "logps/rejected": -13.222760200500488, + "loss": 0.7442, + "rewards/accuracies": 0.78125, + "rewards/chosen": 0.04647160321474075, + "rewards/margins": 1.2347908020019531, + "rewards/rejected": -1.188319206237793, + "step": 828 + }, + { + "epoch": 1.8199780461031834, + "grad_norm": 8.460203170776367, + "learning_rate": 1.223587092621162e-07, + "logits/chosen": -2.316023826599121, + "logits/rejected": -2.315521717071533, + "logps/chosen": -17.78945541381836, + "logps/rejected": -9.205183029174805, + "loss": 0.7188, + "rewards/accuracies": 0.75, + "rewards/chosen": 0.07068800181150436, + "rewards/margins": 0.8467319011688232, + "rewards/rejected": -0.7760438919067383, + "step": 829 + }, + { + "epoch": 1.8221734357848518, + "grad_norm": 14.146021842956543, + "learning_rate": 1.1938028665396172e-07, + "logits/chosen": -2.345604419708252, + "logits/rejected": -2.2585413455963135, + "logps/chosen": -21.479135513305664, + "logps/rejected": -12.763071060180664, + "loss": 0.8413, + "rewards/accuracies": 0.8125, + "rewards/chosen": -0.05769065022468567, + "rewards/margins": 0.865807056427002, + "rewards/rejected": -0.9234977960586548, + "step": 830 + }, + { + "epoch": 1.8243688254665202, + "grad_norm": 4.289247035980225, + "learning_rate": 1.1643767527175858e-07, + "logits/chosen": -2.2771899700164795, + "logits/rejected": -2.280471086502075, + "logps/chosen": -22.956501007080078, + "logps/rejected": -10.655299186706543, + "loss": 0.6877, + "rewards/accuracies": 0.78125, + "rewards/chosen": 0.13649001717567444, + "rewards/margins": 1.2965385913848877, + "rewards/rejected": -1.160048484802246, + "step": 831 + }, + { + "epoch": 1.826564215148189, + "grad_norm": 9.209602355957031, + "learning_rate": 1.1353091938067024e-07, + "logits/chosen": -2.28090238571167, + "logits/rejected": -2.3467183113098145, + "logps/chosen": -20.12339210510254, + "logps/rejected": -10.842389106750488, + "loss": 0.8165, + "rewards/accuracies": 0.6875, + "rewards/chosen": 0.04768510162830353, + "rewards/margins": 0.703514814376831, + "rewards/rejected": -0.6558297276496887, + "step": 832 + }, + { + "epoch": 1.8287596048298573, + "grad_norm": 5.605053901672363, + "learning_rate": 1.1066006270649271e-07, + "logits/chosen": -2.1928863525390625, + "logits/rejected": -2.353243350982666, + "logps/chosen": -24.0958251953125, + "logps/rejected": -7.1851043701171875, + "loss": 0.5952, + "rewards/accuracies": 0.84375, + "rewards/chosen": 0.16257108747959137, + "rewards/margins": 1.4440844058990479, + "rewards/rejected": -1.2815132141113281, + "step": 833 + }, + { + "epoch": 1.8309549945115258, + "grad_norm": 6.619046211242676, + "learning_rate": 1.0782514843499653e-07, + "logits/chosen": -2.3190622329711914, + "logits/rejected": -2.307655096054077, + "logps/chosen": -24.60797119140625, + "logps/rejected": -12.444743156433105, + "loss": 0.7256, + "rewards/accuracies": 0.84375, + "rewards/chosen": 0.11972400546073914, + "rewards/margins": 1.2502537965774536, + "rewards/rejected": -1.1305298805236816, + "step": 834 + }, + { + "epoch": 1.8331503841931944, + "grad_norm": 14.643877029418945, + "learning_rate": 1.0502621921127776e-07, + "logits/chosen": -2.298534870147705, + "logits/rejected": -2.3237557411193848, + "logps/chosen": -23.761741638183594, + "logps/rejected": -9.228866577148438, + "loss": 0.651, + "rewards/accuracies": 0.8125, + "rewards/chosen": 0.04507717490196228, + "rewards/margins": 1.2791664600372314, + "rewards/rejected": -1.2340892553329468, + "step": 835 + }, + { + "epoch": 1.8353457738748626, + "grad_norm": 14.517144203186035, + "learning_rate": 1.0226331713911547e-07, + "logits/chosen": -2.2696497440338135, + "logits/rejected": -2.3429954051971436, + "logps/chosen": -25.594778060913086, + "logps/rejected": -10.700637817382812, + "loss": 0.7019, + "rewards/accuracies": 0.8125, + "rewards/chosen": -0.09721105545759201, + "rewards/margins": 1.3458188772201538, + "rewards/rejected": -1.4430299997329712, + "step": 836 + }, + { + "epoch": 1.8375411635565313, + "grad_norm": 14.31534481048584, + "learning_rate": 9.95364837803392e-08, + "logits/chosen": -2.2340893745422363, + "logits/rejected": -2.2756853103637695, + "logps/chosen": -22.855770111083984, + "logps/rejected": -11.31765079498291, + "loss": 0.7731, + "rewards/accuracies": 0.8125, + "rewards/chosen": 0.044228676706552505, + "rewards/margins": 0.9392744302749634, + "rewards/rejected": -0.8950457572937012, + "step": 837 + }, + { + "epoch": 1.8397365532381997, + "grad_norm": 7.147467613220215, + "learning_rate": 9.684576015420277e-08, + "logits/chosen": -2.2760350704193115, + "logits/rejected": -2.2930829524993896, + "logps/chosen": -24.15088653564453, + "logps/rejected": -11.590227127075195, + "loss": 0.7363, + "rewards/accuracies": 0.78125, + "rewards/chosen": -0.011523757129907608, + "rewards/margins": 1.102854609489441, + "rewards/rejected": -1.1143782138824463, + "step": 838 + }, + { + "epoch": 1.8419319429198682, + "grad_norm": 11.070422172546387, + "learning_rate": 9.419118673676924e-08, + "logits/chosen": -2.3097023963928223, + "logits/rejected": -2.3340721130371094, + "logps/chosen": -26.131114959716797, + "logps/rejected": -16.626916885375977, + "loss": 0.8524, + "rewards/accuracies": 0.71875, + "rewards/chosen": -0.027499038726091385, + "rewards/margins": 0.9677544832229614, + "rewards/rejected": -0.9952535033226013, + "step": 839 + }, + { + "epoch": 1.8441273326015368, + "grad_norm": 9.542285919189453, + "learning_rate": 9.157280346029918e-08, + "logits/chosen": -2.2687957286834717, + "logits/rejected": -2.273031711578369, + "logps/chosen": -20.57872772216797, + "logps/rejected": -16.403339385986328, + "loss": 0.9546, + "rewards/accuracies": 0.78125, + "rewards/chosen": 0.05885916203260422, + "rewards/margins": 0.6920310258865356, + "rewards/rejected": -0.633171796798706, + "step": 840 + }, + { + "epoch": 1.8463227222832053, + "grad_norm": 11.59989070892334, + "learning_rate": 8.899064971265275e-08, + "logits/chosen": -2.2603890895843506, + "logits/rejected": -2.2805569171905518, + "logps/chosen": -21.231781005859375, + "logps/rejected": -12.615869522094727, + "loss": 0.8059, + "rewards/accuracies": 0.78125, + "rewards/chosen": -0.0033099502325057983, + "rewards/margins": 0.798790693283081, + "rewards/rejected": -0.8021005392074585, + "step": 841 + }, + { + "epoch": 1.8485181119648737, + "grad_norm": 7.000414848327637, + "learning_rate": 8.64447643366953e-08, + "logits/chosen": -2.2215676307678223, + "logits/rejected": -2.3573269844055176, + "logps/chosen": -21.28950309753418, + "logps/rejected": -11.500212669372559, + "loss": 0.7886, + "rewards/accuracies": 0.78125, + "rewards/chosen": 0.012604400515556335, + "rewards/margins": 0.9372907876968384, + "rewards/rejected": -0.9246863126754761, + "step": 842 + }, + { + "epoch": 1.8507135016465424, + "grad_norm": 6.195981979370117, + "learning_rate": 8.393518562971353e-08, + "logits/chosen": -2.3180973529815674, + "logits/rejected": -2.354137897491455, + "logps/chosen": -19.93052101135254, + "logps/rejected": -9.889596939086914, + "loss": 0.7036, + "rewards/accuracies": 0.71875, + "rewards/chosen": 0.1700046956539154, + "rewards/margins": 1.0228936672210693, + "rewards/rejected": -0.8528889417648315, + "step": 843 + }, + { + "epoch": 1.8529088913282108, + "grad_norm": 5.888799667358398, + "learning_rate": 8.146195134284052e-08, + "logits/chosen": -2.2787113189697266, + "logits/rejected": -2.284717559814453, + "logps/chosen": -27.794681549072266, + "logps/rejected": -11.615095138549805, + "loss": 0.6483, + "rewards/accuracies": 0.90625, + "rewards/chosen": 0.011686277575790882, + "rewards/margins": 1.444423794746399, + "rewards/rejected": -1.4327374696731567, + "step": 844 + }, + { + "epoch": 1.8551042810098792, + "grad_norm": 8.796133041381836, + "learning_rate": 7.902509868048552e-08, + "logits/chosen": -2.30924129486084, + "logits/rejected": -2.2898685932159424, + "logps/chosen": -26.780763626098633, + "logps/rejected": -13.234761238098145, + "loss": 0.7297, + "rewards/accuracies": 0.875, + "rewards/chosen": 0.087000772356987, + "rewards/margins": 1.4246666431427002, + "rewards/rejected": -1.3376657962799072, + "step": 845 + }, + { + "epoch": 1.8572996706915479, + "grad_norm": 16.145828247070312, + "learning_rate": 7.6624664299777e-08, + "logits/chosen": -2.3867878913879395, + "logits/rejected": -2.3249454498291016, + "logps/chosen": -18.916215896606445, + "logps/rejected": -11.309992790222168, + "loss": 0.8776, + "rewards/accuracies": 0.6875, + "rewards/chosen": -0.10765193402767181, + "rewards/margins": 0.6062726378440857, + "rewards/rejected": -0.7139245271682739, + "step": 846 + }, + { + "epoch": 1.859495060373216, + "grad_norm": 14.361079216003418, + "learning_rate": 7.426068431000883e-08, + "logits/chosen": -2.2815101146698, + "logits/rejected": -2.2684407234191895, + "logps/chosen": -28.854141235351562, + "logps/rejected": -13.138442039489746, + "loss": 0.721, + "rewards/accuracies": 0.875, + "rewards/chosen": 0.0018514394760131836, + "rewards/margins": 1.4000728130340576, + "rewards/rejected": -1.3982213735580444, + "step": 847 + }, + { + "epoch": 1.8616904500548848, + "grad_norm": 6.622011661529541, + "learning_rate": 7.19331942720991e-08, + "logits/chosen": -2.3182671070098877, + "logits/rejected": -2.3678479194641113, + "logps/chosen": -18.305288314819336, + "logps/rejected": -8.286081314086914, + "loss": 0.744, + "rewards/accuracies": 0.6875, + "rewards/chosen": 0.10816448926925659, + "rewards/margins": 0.8645082712173462, + "rewards/rejected": -0.7563438415527344, + "step": 848 + }, + { + "epoch": 1.8638858397365532, + "grad_norm": 6.886569023132324, + "learning_rate": 6.964222919805391e-08, + "logits/chosen": -2.255606174468994, + "logits/rejected": -2.2938742637634277, + "logps/chosen": -23.53927993774414, + "logps/rejected": -13.567049980163574, + "loss": 0.8183, + "rewards/accuracies": 0.75, + "rewards/chosen": 0.0012741461396217346, + "rewards/margins": 0.9946119785308838, + "rewards/rejected": -0.993337869644165, + "step": 849 + }, + { + "epoch": 1.8660812294182216, + "grad_norm": 14.821948051452637, + "learning_rate": 6.738782355044048e-08, + "logits/chosen": -2.232952356338501, + "logits/rejected": -2.383232831954956, + "logps/chosen": -25.572799682617188, + "logps/rejected": -11.867044448852539, + "loss": 0.8401, + "rewards/accuracies": 0.75, + "rewards/chosen": -0.1324472278356552, + "rewards/margins": 1.0741969347000122, + "rewards/rejected": -1.2066442966461182, + "step": 850 + }, + { + "epoch": 1.8682766190998903, + "grad_norm": 7.198695182800293, + "learning_rate": 6.51700112418699e-08, + "logits/chosen": -2.247284412384033, + "logits/rejected": -2.3382396697998047, + "logps/chosen": -23.003253936767578, + "logps/rejected": -13.087779998779297, + "loss": 0.7269, + "rewards/accuracies": 0.78125, + "rewards/chosen": -0.056551843881607056, + "rewards/margins": 0.9394438862800598, + "rewards/rejected": -0.9959957599639893, + "step": 851 + }, + { + "epoch": 1.8704720087815587, + "grad_norm": 8.24974250793457, + "learning_rate": 6.2988825634486e-08, + "logits/chosen": -2.3439106941223145, + "logits/rejected": -2.282029390335083, + "logps/chosen": -28.53689956665039, + "logps/rejected": -14.327596664428711, + "loss": 0.8074, + "rewards/accuracies": 0.8125, + "rewards/chosen": -0.01979568600654602, + "rewards/margins": 1.089199185371399, + "rewards/rejected": -1.1089948415756226, + "step": 852 + }, + { + "epoch": 1.8726673984632272, + "grad_norm": 9.166004180908203, + "learning_rate": 6.084429953946452e-08, + "logits/chosen": -2.3002305030822754, + "logits/rejected": -2.3293113708496094, + "logps/chosen": -18.435104370117188, + "logps/rejected": -13.096721649169922, + "loss": 0.8452, + "rewards/accuracies": 0.84375, + "rewards/chosen": -0.006609760224819183, + "rewards/margins": 0.7510573863983154, + "rewards/rejected": -0.7576671242713928, + "step": 853 + }, + { + "epoch": 1.8748627881448958, + "grad_norm": 7.49564266204834, + "learning_rate": 5.87364652165176e-08, + "logits/chosen": -2.312483310699463, + "logits/rejected": -2.36527156829834, + "logps/chosen": -22.692251205444336, + "logps/rejected": -9.761503219604492, + "loss": 0.7306, + "rewards/accuracies": 0.78125, + "rewards/chosen": 0.047489233314991, + "rewards/margins": 0.8913099765777588, + "rewards/rejected": -0.8438207507133484, + "step": 854 + }, + { + "epoch": 1.8770581778265643, + "grad_norm": 8.075494766235352, + "learning_rate": 5.6665354373411085e-08, + "logits/chosen": -2.297520637512207, + "logits/rejected": -2.380192279815674, + "logps/chosen": -20.89752197265625, + "logps/rejected": -9.794967651367188, + "loss": 0.7761, + "rewards/accuracies": 0.84375, + "rewards/chosen": 0.033703841269016266, + "rewards/margins": 0.970736026763916, + "rewards/rejected": -0.937032163143158, + "step": 855 + }, + { + "epoch": 1.8792535675082327, + "grad_norm": 13.935422897338867, + "learning_rate": 5.463099816548578e-08, + "logits/chosen": -2.2718095779418945, + "logits/rejected": -2.2321574687957764, + "logps/chosen": -19.762210845947266, + "logps/rejected": -11.357540130615234, + "loss": 0.763, + "rewards/accuracies": 0.78125, + "rewards/chosen": -0.017398372292518616, + "rewards/margins": 0.9169367551803589, + "rewards/rejected": -0.9343350529670715, + "step": 856 + }, + { + "epoch": 1.8814489571899013, + "grad_norm": 7.167140007019043, + "learning_rate": 5.263342719518921e-08, + "logits/chosen": -2.2818827629089355, + "logits/rejected": -2.2975621223449707, + "logps/chosen": -29.85270118713379, + "logps/rejected": -10.076404571533203, + "loss": 0.6159, + "rewards/accuracies": 0.84375, + "rewards/chosen": 0.06604423373937607, + "rewards/margins": 1.6125178337097168, + "rewards/rejected": -1.5464736223220825, + "step": 857 + }, + { + "epoch": 1.8836443468715696, + "grad_norm": 8.877127647399902, + "learning_rate": 5.0672671511615144e-08, + "logits/chosen": -2.402839183807373, + "logits/rejected": -2.425827741622925, + "logps/chosen": -24.932357788085938, + "logps/rejected": -14.4828519821167, + "loss": 0.9132, + "rewards/accuracies": 0.75, + "rewards/chosen": -0.08172663301229477, + "rewards/margins": 0.850361704826355, + "rewards/rejected": -0.932088315486908, + "step": 858 + }, + { + "epoch": 1.8858397365532382, + "grad_norm": 4.971401691436768, + "learning_rate": 4.8748760610051735e-08, + "logits/chosen": -2.3461244106292725, + "logits/rejected": -2.326270580291748, + "logps/chosen": -23.266752243041992, + "logps/rejected": -11.518921852111816, + "loss": 0.7699, + "rewards/accuracies": 0.78125, + "rewards/chosen": 0.11916793882846832, + "rewards/margins": 0.9273298978805542, + "rewards/rejected": -0.8081620335578918, + "step": 859 + }, + { + "epoch": 1.8880351262349067, + "grad_norm": 8.389633178710938, + "learning_rate": 4.6861723431538273e-08, + "logits/chosen": -2.3503756523132324, + "logits/rejected": -2.323322057723999, + "logps/chosen": -22.747604370117188, + "logps/rejected": -14.256501197814941, + "loss": 0.877, + "rewards/accuracies": 0.65625, + "rewards/chosen": -0.13185255229473114, + "rewards/margins": 0.8785985708236694, + "rewards/rejected": -1.0104511976242065, + "step": 860 + }, + { + "epoch": 1.890230515916575, + "grad_norm": 11.095335006713867, + "learning_rate": 4.501158836242914e-08, + "logits/chosen": -2.263857126235962, + "logits/rejected": -2.3446054458618164, + "logps/chosen": -19.456649780273438, + "logps/rejected": -10.87272834777832, + "loss": 0.8247, + "rewards/accuracies": 0.84375, + "rewards/chosen": 0.0570753738284111, + "rewards/margins": 0.698641836643219, + "rewards/rejected": -0.6415663957595825, + "step": 861 + }, + { + "epoch": 1.8924259055982438, + "grad_norm": 7.88322639465332, + "learning_rate": 4.319838323396691e-08, + "logits/chosen": -2.353119373321533, + "logits/rejected": -2.2819082736968994, + "logps/chosen": -31.483495712280273, + "logps/rejected": -8.964349746704102, + "loss": 0.5743, + "rewards/accuracies": 0.875, + "rewards/chosen": 0.11409927904605865, + "rewards/margins": 1.3078382015228271, + "rewards/rejected": -1.1937389373779297, + "step": 862 + }, + { + "epoch": 1.8946212952799122, + "grad_norm": 6.945970058441162, + "learning_rate": 4.142213532186384e-08, + "logits/chosen": -2.283721446990967, + "logits/rejected": -2.3039913177490234, + "logps/chosen": -22.71794891357422, + "logps/rejected": -8.919666290283203, + "loss": 0.6486, + "rewards/accuracies": 0.84375, + "rewards/chosen": 0.09672074019908905, + "rewards/margins": 1.1749136447906494, + "rewards/rejected": -1.078192949295044, + "step": 863 + }, + { + "epoch": 1.8968166849615806, + "grad_norm": 18.7995548248291, + "learning_rate": 3.968287134589188e-08, + "logits/chosen": -2.2822353839874268, + "logits/rejected": -2.343181848526001, + "logps/chosen": -24.483694076538086, + "logps/rejected": -11.352913856506348, + "loss": 0.6599, + "rewards/accuracies": 0.90625, + "rewards/chosen": 0.052551690489053726, + "rewards/margins": 1.2649013996124268, + "rewards/rejected": -1.2123498916625977, + "step": 864 + }, + { + "epoch": 1.8990120746432493, + "grad_norm": 6.598869800567627, + "learning_rate": 3.798061746947995e-08, + "logits/chosen": -2.343810558319092, + "logits/rejected": -2.263450860977173, + "logps/chosen": -24.16651725769043, + "logps/rejected": -12.646990776062012, + "loss": 0.7594, + "rewards/accuracies": 0.78125, + "rewards/chosen": 0.06020158529281616, + "rewards/margins": 1.1387993097305298, + "rewards/rejected": -1.0785977840423584, + "step": 865 + }, + { + "epoch": 1.9012074643249177, + "grad_norm": 8.282346725463867, + "learning_rate": 3.631539929932149e-08, + "logits/chosen": -2.2230026721954346, + "logits/rejected": -2.1961493492126465, + "logps/chosen": -21.13140869140625, + "logps/rejected": -10.19251823425293, + "loss": 0.6738, + "rewards/accuracies": 0.8125, + "rewards/chosen": 0.23780131340026855, + "rewards/margins": 1.1513222455978394, + "rewards/rejected": -0.9135209321975708, + "step": 866 + }, + { + "epoch": 1.9034028540065862, + "grad_norm": 8.047040939331055, + "learning_rate": 3.468724188498751e-08, + "logits/chosen": -2.1813344955444336, + "logits/rejected": -2.3277149200439453, + "logps/chosen": -19.10706329345703, + "logps/rejected": -10.097394943237305, + "loss": 0.7673, + "rewards/accuracies": 0.78125, + "rewards/chosen": -0.006772905588150024, + "rewards/margins": 0.9838821291923523, + "rewards/rejected": -0.9906550645828247, + "step": 867 + }, + { + "epoch": 1.9055982436882548, + "grad_norm": 7.088027477264404, + "learning_rate": 3.309616971855195e-08, + "logits/chosen": -2.230976104736328, + "logits/rejected": -2.3410205841064453, + "logps/chosen": -27.349727630615234, + "logps/rejected": -13.83423137664795, + "loss": 0.8491, + "rewards/accuracies": 0.71875, + "rewards/chosen": 0.013154106214642525, + "rewards/margins": 1.0419939756393433, + "rewards/rejected": -1.0288399457931519, + "step": 868 + }, + { + "epoch": 1.907793633369923, + "grad_norm": 9.376320838928223, + "learning_rate": 3.154220673422193e-08, + "logits/chosen": -2.338898181915283, + "logits/rejected": -2.325963258743286, + "logps/chosen": -23.3134708404541, + "logps/rejected": -10.624837875366211, + "loss": 0.7366, + "rewards/accuracies": 0.78125, + "rewards/chosen": -0.002025183290243149, + "rewards/margins": 1.0471056699752808, + "rewards/rejected": -1.049130916595459, + "step": 869 + }, + { + "epoch": 1.9099890230515917, + "grad_norm": 8.06198501586914, + "learning_rate": 3.0025376307977474e-08, + "logits/chosen": -2.2715907096862793, + "logits/rejected": -2.2343966960906982, + "logps/chosen": -24.380752563476562, + "logps/rejected": -11.601836204528809, + "loss": 0.7286, + "rewards/accuracies": 0.84375, + "rewards/chosen": 0.1475653201341629, + "rewards/margins": 1.0858004093170166, + "rewards/rejected": -0.9382350444793701, + "step": 870 + }, + { + "epoch": 1.9121844127332601, + "grad_norm": 7.233065128326416, + "learning_rate": 2.8545701257221003e-08, + "logits/chosen": -2.2290115356445312, + "logits/rejected": -2.340165138244629, + "logps/chosen": -28.525609970092773, + "logps/rejected": -14.161710739135742, + "loss": 0.7588, + "rewards/accuracies": 0.9375, + "rewards/chosen": 0.09254666417837143, + "rewards/margins": 1.2745497226715088, + "rewards/rejected": -1.182003140449524, + "step": 871 + }, + { + "epoch": 1.9143798024149286, + "grad_norm": 12.206696510314941, + "learning_rate": 2.71032038404323e-08, + "logits/chosen": -2.3242523670196533, + "logits/rejected": -2.339348316192627, + "logps/chosen": -18.249536514282227, + "logps/rejected": -9.425407409667969, + "loss": 0.7992, + "rewards/accuracies": 0.6875, + "rewards/chosen": 0.050805263221263885, + "rewards/margins": 0.5615851879119873, + "rewards/rejected": -0.5107799172401428, + "step": 872 + }, + { + "epoch": 1.9165751920965972, + "grad_norm": 7.801868438720703, + "learning_rate": 2.569790575683656e-08, + "logits/chosen": -2.2905728816986084, + "logits/rejected": -2.28438663482666, + "logps/chosen": -21.798429489135742, + "logps/rejected": -11.017197608947754, + "loss": 0.7771, + "rewards/accuracies": 0.8125, + "rewards/chosen": 0.020733946934342384, + "rewards/margins": 0.9085362553596497, + "rewards/rejected": -0.887802243232727, + "step": 873 + }, + { + "epoch": 1.9187705817782656, + "grad_norm": 15.521644592285156, + "learning_rate": 2.4329828146074096e-08, + "logits/chosen": -2.2940855026245117, + "logits/rejected": -2.2680234909057617, + "logps/chosen": -18.777267456054688, + "logps/rejected": -10.669624328613281, + "loss": 0.8116, + "rewards/accuracies": 0.75, + "rewards/chosen": 0.09511300921440125, + "rewards/margins": 0.7384185791015625, + "rewards/rejected": -0.6433055996894836, + "step": 874 + }, + { + "epoch": 1.920965971459934, + "grad_norm": 6.091784477233887, + "learning_rate": 2.299899158788671e-08, + "logits/chosen": -2.294455051422119, + "logits/rejected": -2.2514615058898926, + "logps/chosen": -21.79439353942871, + "logps/rejected": -10.755569458007812, + "loss": 0.7313, + "rewards/accuracies": 0.78125, + "rewards/chosen": 0.07072372734546661, + "rewards/margins": 0.9454042911529541, + "rewards/rejected": -0.8746806383132935, + "step": 875 + }, + { + "epoch": 1.9231613611416027, + "grad_norm": 17.321889877319336, + "learning_rate": 2.1705416101804323e-08, + "logits/chosen": -2.364165782928467, + "logits/rejected": -2.269895076751709, + "logps/chosen": -17.51275062561035, + "logps/rejected": -10.912698745727539, + "loss": 0.8682, + "rewards/accuracies": 0.5625, + "rewards/chosen": -0.1837128847837448, + "rewards/margins": 0.537417471408844, + "rewards/rejected": -0.7211303114891052, + "step": 876 + }, + { + "epoch": 1.9253567508232712, + "grad_norm": 6.5813889503479, + "learning_rate": 2.0449121146845773e-08, + "logits/chosen": -2.2546608448028564, + "logits/rejected": -2.2185187339782715, + "logps/chosen": -22.255146026611328, + "logps/rejected": -11.314976692199707, + "loss": 0.7552, + "rewards/accuracies": 0.8125, + "rewards/chosen": 0.050574660301208496, + "rewards/margins": 1.0326803922653198, + "rewards/rejected": -0.9821057319641113, + "step": 877 + }, + { + "epoch": 1.9275521405049396, + "grad_norm": 8.453032493591309, + "learning_rate": 1.9230125621225727e-08, + "logits/chosen": -2.280209541320801, + "logits/rejected": -2.3256332874298096, + "logps/chosen": -23.86482810974121, + "logps/rejected": -8.854063034057617, + "loss": 0.643, + "rewards/accuracies": 0.84375, + "rewards/chosen": 0.06025511771440506, + "rewards/margins": 1.2743371725082397, + "rewards/rejected": -1.2140820026397705, + "step": 878 + }, + { + "epoch": 1.9297475301866083, + "grad_norm": 8.441838264465332, + "learning_rate": 1.8048447862070718e-08, + "logits/chosen": -2.2905054092407227, + "logits/rejected": -2.2417261600494385, + "logps/chosen": -20.974136352539062, + "logps/rejected": -10.553691864013672, + "loss": 0.7324, + "rewards/accuracies": 0.78125, + "rewards/chosen": 0.05540268123149872, + "rewards/margins": 0.9626655578613281, + "rewards/rejected": -0.907262921333313, + "step": 879 + }, + { + "epoch": 1.9319429198682765, + "grad_norm": 13.303086280822754, + "learning_rate": 1.6904105645142443e-08, + "logits/chosen": -2.308682680130005, + "logits/rejected": -2.3087410926818848, + "logps/chosen": -22.469852447509766, + "logps/rejected": -12.35126781463623, + "loss": 0.7576, + "rewards/accuracies": 0.875, + "rewards/chosen": 0.018310165032744408, + "rewards/margins": 1.1040092706680298, + "rewards/rejected": -1.0856990814208984, + "step": 880 + }, + { + "epoch": 1.9341383095499451, + "grad_norm": 8.383279800415039, + "learning_rate": 1.5797116184571305e-08, + "logits/chosen": -2.283200740814209, + "logits/rejected": -2.280250072479248, + "logps/chosen": -25.64556884765625, + "logps/rejected": -11.530251502990723, + "loss": 0.7423, + "rewards/accuracies": 0.78125, + "rewards/chosen": 0.028892913833260536, + "rewards/margins": 1.0426300764083862, + "rewards/rejected": -1.0137372016906738, + "step": 881 + }, + { + "epoch": 1.9363336992316136, + "grad_norm": 13.752285957336426, + "learning_rate": 1.4727496132596608e-08, + "logits/chosen": -2.3798322677612305, + "logits/rejected": -2.371227741241455, + "logps/chosen": -28.240768432617188, + "logps/rejected": -13.266584396362305, + "loss": 0.8034, + "rewards/accuracies": 0.78125, + "rewards/chosen": -0.009360112249851227, + "rewards/margins": 1.168959617614746, + "rewards/rejected": -1.1783196926116943, + "step": 882 + }, + { + "epoch": 1.938529088913282, + "grad_norm": 9.240599632263184, + "learning_rate": 1.3695261579316776e-08, + "logits/chosen": -2.272313356399536, + "logits/rejected": -2.267460584640503, + "logps/chosen": -20.466670989990234, + "logps/rejected": -9.543893814086914, + "loss": 0.7226, + "rewards/accuracies": 0.84375, + "rewards/chosen": 0.0357314869761467, + "rewards/margins": 1.0523828268051147, + "rewards/rejected": -1.0166513919830322, + "step": 883 + }, + { + "epoch": 1.9407244785949507, + "grad_norm": 8.774166107177734, + "learning_rate": 1.2700428052447033e-08, + "logits/chosen": -2.2724432945251465, + "logits/rejected": -2.259981393814087, + "logps/chosen": -18.321439743041992, + "logps/rejected": -10.12303638458252, + "loss": 0.7676, + "rewards/accuracies": 0.71875, + "rewards/chosen": -0.012789766304194927, + "rewards/margins": 0.7662094831466675, + "rewards/rejected": -0.7789992094039917, + "step": 884 + }, + { + "epoch": 1.942919868276619, + "grad_norm": 12.818648338317871, + "learning_rate": 1.1743010517085428e-08, + "logits/chosen": -2.3885865211486816, + "logits/rejected": -2.429250478744507, + "logps/chosen": -22.339385986328125, + "logps/rejected": -10.031664848327637, + "loss": 0.8093, + "rewards/accuracies": 0.78125, + "rewards/chosen": -0.043203018605709076, + "rewards/margins": 0.9644555449485779, + "rewards/rejected": -1.00765860080719, + "step": 885 + }, + { + "epoch": 1.9451152579582875, + "grad_norm": 4.587555885314941, + "learning_rate": 1.0823023375489128e-08, + "logits/chosen": -2.3121843338012695, + "logits/rejected": -2.412358283996582, + "logps/chosen": -21.93047332763672, + "logps/rejected": -9.082547187805176, + "loss": 0.6474, + "rewards/accuracies": 0.90625, + "rewards/chosen": 0.1289997398853302, + "rewards/margins": 1.1321951150894165, + "rewards/rejected": -1.0031952857971191, + "step": 886 + }, + { + "epoch": 1.9473106476399562, + "grad_norm": 7.059382438659668, + "learning_rate": 9.940480466855418e-09, + "logits/chosen": -2.2519495487213135, + "logits/rejected": -2.2558467388153076, + "logps/chosen": -27.83672332763672, + "logps/rejected": -10.076109886169434, + "loss": 0.6779, + "rewards/accuracies": 0.8125, + "rewards/chosen": 0.1005638986825943, + "rewards/margins": 1.2121317386627197, + "rewards/rejected": -1.1115678548812866, + "step": 887 + }, + { + "epoch": 1.9495060373216246, + "grad_norm": 8.29926586151123, + "learning_rate": 9.095395067115764e-09, + "logits/chosen": -2.2163219451904297, + "logits/rejected": -2.3477697372436523, + "logps/chosen": -22.68077278137207, + "logps/rejected": -10.835956573486328, + "loss": 0.7158, + "rewards/accuracies": 0.84375, + "rewards/chosen": 0.107290118932724, + "rewards/margins": 1.0432050228118896, + "rewards/rejected": -0.9359148740768433, + "step": 888 + }, + { + "epoch": 1.951701427003293, + "grad_norm": 10.054495811462402, + "learning_rate": 8.28777988873486e-09, + "logits/chosen": -2.295466899871826, + "logits/rejected": -2.3601460456848145, + "logps/chosen": -21.413726806640625, + "logps/rejected": -11.572498321533203, + "loss": 0.8269, + "rewards/accuracies": 0.78125, + "rewards/chosen": -0.04080420359969139, + "rewards/margins": 0.722064197063446, + "rewards/rejected": -0.7628684043884277, + "step": 889 + }, + { + "epoch": 1.9538968166849617, + "grad_norm": 9.413023948669434, + "learning_rate": 7.517647080519941e-09, + "logits/chosen": -2.3161637783050537, + "logits/rejected": -2.3140461444854736, + "logps/chosen": -28.21819305419922, + "logps/rejected": -11.886323928833008, + "loss": 0.7061, + "rewards/accuracies": 0.8125, + "rewards/chosen": -0.016428498551249504, + "rewards/margins": 1.2815443277359009, + "rewards/rejected": -1.2979729175567627, + "step": 890 + }, + { + "epoch": 1.95609220636663, + "grad_norm": 11.771985054016113, + "learning_rate": 6.785008227437329e-09, + "logits/chosen": -2.2918152809143066, + "logits/rejected": -2.2907934188842773, + "logps/chosen": -21.773279190063477, + "logps/rejected": -10.203116416931152, + "loss": 0.8177, + "rewards/accuracies": 0.75, + "rewards/chosen": -0.023505816236138344, + "rewards/margins": 0.977125883102417, + "rewards/rejected": -1.0006318092346191, + "step": 891 + }, + { + "epoch": 1.9582875960482986, + "grad_norm": 10.504355430603027, + "learning_rate": 6.089874350439507e-09, + "logits/chosen": -2.320657730102539, + "logits/rejected": -2.351315498352051, + "logps/chosen": -22.519603729248047, + "logps/rejected": -10.951231002807617, + "loss": 0.7281, + "rewards/accuracies": 0.8125, + "rewards/chosen": 0.20374858379364014, + "rewards/margins": 1.07362961769104, + "rewards/rejected": -0.8698810338973999, + "step": 892 + }, + { + "epoch": 1.960482985729967, + "grad_norm": 6.943790435791016, + "learning_rate": 5.432255906297479e-09, + "logits/chosen": -2.3118720054626465, + "logits/rejected": -2.2375969886779785, + "logps/chosen": -24.68213653564453, + "logps/rejected": -11.798558235168457, + "loss": 0.6462, + "rewards/accuracies": 0.875, + "rewards/chosen": 0.15215079486370087, + "rewards/margins": 1.296844244003296, + "rewards/rejected": -1.144693374633789, + "step": 893 + }, + { + "epoch": 1.9626783754116355, + "grad_norm": 11.376214981079102, + "learning_rate": 4.812162787445063e-09, + "logits/chosen": -2.3501124382019043, + "logits/rejected": -2.2765772342681885, + "logps/chosen": -24.22418212890625, + "logps/rejected": -12.231316566467285, + "loss": 0.7663, + "rewards/accuracies": 0.6875, + "rewards/chosen": -0.06302346289157867, + "rewards/margins": 1.0231250524520874, + "rewards/rejected": -1.0861485004425049, + "step": 894 + }, + { + "epoch": 1.9648737650933041, + "grad_norm": 9.542922019958496, + "learning_rate": 4.229604321829561e-09, + "logits/chosen": -2.3719656467437744, + "logits/rejected": -2.310542106628418, + "logps/chosen": -29.604835510253906, + "logps/rejected": -14.17624282836914, + "loss": 0.8143, + "rewards/accuracies": 0.75, + "rewards/chosen": 0.002622615545988083, + "rewards/margins": 1.1476988792419434, + "rewards/rejected": -1.1450762748718262, + "step": 895 + }, + { + "epoch": 1.9670691547749726, + "grad_norm": 12.531002044677734, + "learning_rate": 3.684589272771044e-09, + "logits/chosen": -2.2977163791656494, + "logits/rejected": -2.3711047172546387, + "logps/chosen": -20.241180419921875, + "logps/rejected": -8.573559761047363, + "loss": 0.6766, + "rewards/accuracies": 0.71875, + "rewards/chosen": 0.07672523707151413, + "rewards/margins": 0.8753741383552551, + "rewards/rejected": -0.7986489534378052, + "step": 896 + }, + { + "epoch": 1.969264544456641, + "grad_norm": 7.01442289352417, + "learning_rate": 3.177125838830786e-09, + "logits/chosen": -2.3161184787750244, + "logits/rejected": -2.319521427154541, + "logps/chosen": -24.766826629638672, + "logps/rejected": -8.103300094604492, + "loss": 0.5916, + "rewards/accuracies": 0.90625, + "rewards/chosen": 0.1568806916475296, + "rewards/margins": 1.3657246828079224, + "rewards/rejected": -1.2088439464569092, + "step": 897 + }, + { + "epoch": 1.9714599341383097, + "grad_norm": 9.482521057128906, + "learning_rate": 2.7072216536885855e-09, + "logits/chosen": -2.2606639862060547, + "logits/rejected": -2.3403496742248535, + "logps/chosen": -27.123733520507812, + "logps/rejected": -12.219449043273926, + "loss": 0.7528, + "rewards/accuracies": 0.8125, + "rewards/chosen": -0.003182481974363327, + "rewards/margins": 1.303727388381958, + "rewards/rejected": -1.3069097995758057, + "step": 898 + }, + { + "epoch": 1.9736553238199779, + "grad_norm": 7.0323872566223145, + "learning_rate": 2.2748837860270267e-09, + "logits/chosen": -2.3276515007019043, + "logits/rejected": -2.3275508880615234, + "logps/chosen": -23.31214141845703, + "logps/rejected": -11.630393028259277, + "loss": 0.7581, + "rewards/accuracies": 0.75, + "rewards/chosen": 0.06022808700799942, + "rewards/margins": 1.0111446380615234, + "rewards/rejected": -0.9509164094924927, + "step": 899 + }, + { + "epoch": 1.9758507135016465, + "grad_norm": 5.9556965827941895, + "learning_rate": 1.8801187394248966e-09, + "logits/chosen": -2.3519766330718994, + "logits/rejected": -2.306118965148926, + "logps/chosen": -24.298009872436523, + "logps/rejected": -9.510473251342773, + "loss": 0.748, + "rewards/accuracies": 0.8125, + "rewards/chosen": 0.12927918136119843, + "rewards/margins": 1.192124605178833, + "rewards/rejected": -1.0628454685211182, + "step": 900 + }, + { + "epoch": 1.9780461031833152, + "grad_norm": 9.604037284851074, + "learning_rate": 1.5229324522605949e-09, + "logits/chosen": -2.3725297451019287, + "logits/rejected": -2.331653594970703, + "logps/chosen": -23.39224624633789, + "logps/rejected": -9.769168853759766, + "loss": 0.6467, + "rewards/accuracies": 0.75, + "rewards/chosen": 0.08794993162155151, + "rewards/margins": 0.9945685863494873, + "rewards/rejected": -0.9066186547279358, + "step": 901 + }, + { + "epoch": 1.9802414928649834, + "grad_norm": 9.361015319824219, + "learning_rate": 1.2033302976222072e-09, + "logits/chosen": -2.314908742904663, + "logits/rejected": -2.2613635063171387, + "logps/chosen": -24.224586486816406, + "logps/rejected": -12.583452224731445, + "loss": 0.7499, + "rewards/accuracies": 0.84375, + "rewards/chosen": 0.02966010943055153, + "rewards/margins": 1.0395889282226562, + "rewards/rejected": -1.009928822517395, + "step": 902 + }, + { + "epoch": 1.982436882546652, + "grad_norm": 12.831734657287598, + "learning_rate": 9.213170832264584e-10, + "logits/chosen": -2.3203463554382324, + "logits/rejected": -2.3400824069976807, + "logps/chosen": -19.536672592163086, + "logps/rejected": -12.142282485961914, + "loss": 0.897, + "rewards/accuracies": 0.8125, + "rewards/chosen": -0.12846747040748596, + "rewards/margins": 0.7801587581634521, + "rewards/rejected": -0.9086262583732605, + "step": 903 + }, + { + "epoch": 1.9846322722283205, + "grad_norm": 8.23582935333252, + "learning_rate": 6.768970513457151e-10, + "logits/chosen": -2.352257013320923, + "logits/rejected": -2.29923152923584, + "logps/chosen": -19.161014556884766, + "logps/rejected": -11.01200008392334, + "loss": 0.7911, + "rewards/accuracies": 0.6875, + "rewards/chosen": 0.025491155683994293, + "rewards/margins": 0.7692648768424988, + "rewards/rejected": -0.7437736988067627, + "step": 904 + }, + { + "epoch": 1.986827661909989, + "grad_norm": 7.7893595695495605, + "learning_rate": 4.700738787466463e-10, + "logits/chosen": -2.368873119354248, + "logits/rejected": -2.2904388904571533, + "logps/chosen": -23.453601837158203, + "logps/rejected": -13.198934555053711, + "loss": 0.8289, + "rewards/accuracies": 0.71875, + "rewards/chosen": 0.02025797963142395, + "rewards/margins": 0.8744856715202332, + "rewards/rejected": -0.854227602481842, + "step": 905 + }, + { + "epoch": 1.9890230515916576, + "grad_norm": 9.57486629486084, + "learning_rate": 3.008506766313812e-10, + "logits/chosen": -2.2960574626922607, + "logits/rejected": -2.2652554512023926, + "logps/chosen": -23.81208038330078, + "logps/rejected": -12.908476829528809, + "loss": 0.753, + "rewards/accuracies": 0.84375, + "rewards/chosen": 0.11780311167240143, + "rewards/margins": 0.950643002986908, + "rewards/rejected": -0.8328399658203125, + "step": 906 + }, + { + "epoch": 1.991218441273326, + "grad_norm": 10.122781753540039, + "learning_rate": 1.692299905944883e-10, + "logits/chosen": -2.351144313812256, + "logits/rejected": -2.334829807281494, + "logps/chosen": -22.798965454101562, + "logps/rejected": -12.726913452148438, + "loss": 0.7387, + "rewards/accuracies": 0.875, + "rewards/chosen": 0.011832953430712223, + "rewards/margins": 0.9553701877593994, + "rewards/rejected": -0.943537175655365, + "step": 907 + }, + { + "epoch": 1.9934138309549945, + "grad_norm": 6.905821800231934, + "learning_rate": 7.52138005810643e-11, + "logits/chosen": -2.289172649383545, + "logits/rejected": -2.232004165649414, + "logps/chosen": -22.383073806762695, + "logps/rejected": -9.590481758117676, + "loss": 0.664, + "rewards/accuracies": 0.8125, + "rewards/chosen": 0.09608496725559235, + "rewards/margins": 1.0758891105651855, + "rewards/rejected": -0.9798040986061096, + "step": 908 + }, + { + "epoch": 1.9956092206366631, + "grad_norm": 8.259413719177246, + "learning_rate": 1.880352085981141e-11, + "logits/chosen": -2.2615325450897217, + "logits/rejected": -2.336507558822632, + "logps/chosen": -21.350929260253906, + "logps/rejected": -10.036890029907227, + "loss": 0.7328, + "rewards/accuracies": 0.875, + "rewards/chosen": 0.10890071839094162, + "rewards/margins": 1.1028611660003662, + "rewards/rejected": -0.9939604997634888, + "step": 909 + }, + { + "epoch": 1.9978046103183313, + "grad_norm": 8.452844619750977, + "learning_rate": 0.0, + "logits/chosen": -2.303891658782959, + "logits/rejected": -2.244595527648926, + "logps/chosen": -22.016010284423828, + "logps/rejected": -10.462875366210938, + "loss": 0.7939, + "rewards/accuracies": 0.78125, + "rewards/chosen": -0.012440510094165802, + "rewards/margins": 1.0303937196731567, + "rewards/rejected": -1.0428342819213867, + "step": 910 + } + ], + "logging_steps": 1.0, + "max_steps": 910, + "num_input_tokens_seen": 0, + "num_train_epochs": 2, + "save_steps": 455, + "stateful_callbacks": { + "TrainerControl": { + "args": { + "should_epoch_stop": false, + "should_evaluate": false, + "should_log": false, + "should_save": true, + "should_training_stop": true + }, + "attributes": {} + } + }, + "total_flos": 7.789415151882469e+18, + "train_batch_size": 2, + "trial_name": null, + "trial_params": null +}