diff --git "a/last-checkpoint/trainer_state.json" "b/last-checkpoint/trainer_state.json" --- "a/last-checkpoint/trainer_state.json" +++ "b/last-checkpoint/trainer_state.json" @@ -1,9 +1,9 @@ { "best_metric": null, "best_model_checkpoint": null, - "epoch": 0.3420996365191362, + "epoch": 1.0262989095574087, "eval_steps": 100, - "global_step": 300, + "global_step": 900, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, @@ -2305,6 +2305,4602 @@ "eval_samples_per_second": 4.575, "eval_steps_per_second": 1.144, "step": 300 + }, + { + "epoch": 0.3443803007625971, + "grad_norm": 84.25578372291146, + "learning_rate": 1.927120518139144e-07, + "logits/chosen": -1.3597999811172485, + "logits/rejected": -1.4241136312484741, + "logps/chosen": -180.1940460205078, + "logps/rejected": -199.74412536621094, + "loss": 0.6421, + "rewards/accuracies": 0.59375, + "rewards/chosen": -0.04787690192461014, + "rewards/margins": 0.13789615035057068, + "rewards/rejected": -0.18577302992343903, + "step": 302 + }, + { + "epoch": 0.34666096500605803, + "grad_norm": 89.24528063714898, + "learning_rate": 1.925688448195983e-07, + "logits/chosen": -1.1900302171707153, + "logits/rejected": -1.3024637699127197, + "logps/chosen": -130.7202606201172, + "logps/rejected": -170.3651123046875, + "loss": 0.6326, + "rewards/accuracies": 0.71875, + "rewards/chosen": 0.08976805210113525, + "rewards/margins": 0.30762431025505066, + "rewards/rejected": -0.2178562879562378, + "step": 304 + }, + { + "epoch": 0.34894162924951894, + "grad_norm": 86.12361588594867, + "learning_rate": 1.924242987531517e-07, + "logits/chosen": -1.3129158020019531, + "logits/rejected": -1.2686495780944824, + "logps/chosen": -170.01808166503906, + "logps/rejected": -172.82835388183594, + "loss": 0.6621, + "rewards/accuracies": 0.59375, + "rewards/chosen": -0.010694226250052452, + "rewards/margins": 0.25143933296203613, + "rewards/rejected": -0.26213353872299194, + "step": 306 + }, + { + "epoch": 0.35122229349297984, + "grad_norm": 80.23753458776699, + "learning_rate": 1.922784157055331e-07, + "logits/chosen": -1.3715269565582275, + "logits/rejected": -1.3273383378982544, + "logps/chosen": -190.35610961914062, + "logps/rejected": -180.68020629882812, + "loss": 0.6522, + "rewards/accuracies": 0.625, + "rewards/chosen": -0.09160695225000381, + "rewards/margins": 0.13991902768611908, + "rewards/rejected": -0.2315259724855423, + "step": 308 + }, + { + "epoch": 0.35350295773644075, + "grad_norm": 98.71267017165292, + "learning_rate": 1.9213119778704127e-07, + "logits/chosen": -1.3962262868881226, + "logits/rejected": -1.4428459405899048, + "logps/chosen": -222.09481811523438, + "logps/rejected": -248.44686889648438, + "loss": 0.6602, + "rewards/accuracies": 0.78125, + "rewards/chosen": -0.001766495406627655, + "rewards/margins": 0.569767415523529, + "rewards/rejected": -0.5715339183807373, + "step": 310 + }, + { + "epoch": 0.35578362197990165, + "grad_norm": 88.65548046490639, + "learning_rate": 1.919826471272849e-07, + "logits/chosen": -1.3152942657470703, + "logits/rejected": -1.365240216255188, + "logps/chosen": -128.6586151123047, + "logps/rejected": -141.93685913085938, + "loss": 0.6252, + "rewards/accuracies": 0.59375, + "rewards/chosen": -0.05175452306866646, + "rewards/margins": 0.1226281225681305, + "rewards/rejected": -0.17438265681266785, + "step": 312 + }, + { + "epoch": 0.35806428622336256, + "grad_norm": 84.35369248301531, + "learning_rate": 1.9183276587515165e-07, + "logits/chosen": -1.2944692373275757, + "logits/rejected": -1.3323063850402832, + "logps/chosen": -165.85975646972656, + "logps/rejected": -191.24041748046875, + "loss": 0.6487, + "rewards/accuracies": 0.53125, + "rewards/chosen": -0.07458823919296265, + "rewards/margins": 0.3784411549568176, + "rewards/rejected": -0.4530293941497803, + "step": 314 + }, + { + "epoch": 0.36034495046682347, + "grad_norm": 86.49341842550122, + "learning_rate": 1.9168155619877707e-07, + "logits/chosen": -1.2946043014526367, + "logits/rejected": -1.4176356792449951, + "logps/chosen": -130.90985107421875, + "logps/rejected": -162.9954071044922, + "loss": 0.6467, + "rewards/accuracies": 0.78125, + "rewards/chosen": 0.011317873373627663, + "rewards/margins": 0.4069710373878479, + "rewards/rejected": -0.39565321803092957, + "step": 316 + }, + { + "epoch": 0.36262561471028437, + "grad_norm": 102.96541053849566, + "learning_rate": 1.9152902028551335e-07, + "logits/chosen": -1.3063163757324219, + "logits/rejected": -1.3172858953475952, + "logps/chosen": -157.0022735595703, + "logps/rejected": -165.20347595214844, + "loss": 0.6669, + "rewards/accuracies": 0.59375, + "rewards/chosen": -0.07298893481492996, + "rewards/margins": 0.17434003949165344, + "rewards/rejected": -0.2473289668560028, + "step": 318 + }, + { + "epoch": 0.3649062789537453, + "grad_norm": 76.39528252287091, + "learning_rate": 1.9137516034189765e-07, + "logits/chosen": -1.3794448375701904, + "logits/rejected": -1.4233444929122925, + "logps/chosen": -164.51431274414062, + "logps/rejected": -191.4024658203125, + "loss": 0.629, + "rewards/accuracies": 0.71875, + "rewards/chosen": 0.07598446309566498, + "rewards/margins": 0.355991005897522, + "rewards/rejected": -0.2800065577030182, + "step": 320 + }, + { + "epoch": 0.3671869431972062, + "grad_norm": 90.58917607125439, + "learning_rate": 1.9121997859361997e-07, + "logits/chosen": -1.3726584911346436, + "logits/rejected": -1.4218438863754272, + "logps/chosen": -161.8759765625, + "logps/rejected": -191.62745666503906, + "loss": 0.6777, + "rewards/accuracies": 0.59375, + "rewards/chosen": -0.03798535838723183, + "rewards/margins": 0.2353454828262329, + "rewards/rejected": -0.27333083748817444, + "step": 322 + }, + { + "epoch": 0.3694676074406671, + "grad_norm": 88.70829690205912, + "learning_rate": 1.9106347728549132e-07, + "logits/chosen": -1.4393202066421509, + "logits/rejected": -1.4652281999588013, + "logps/chosen": -159.0114288330078, + "logps/rejected": -169.2324981689453, + "loss": 0.6457, + "rewards/accuracies": 0.59375, + "rewards/chosen": -0.03970104455947876, + "rewards/margins": 0.19649550318717957, + "rewards/rejected": -0.23619654774665833, + "step": 324 + }, + { + "epoch": 0.371748271684128, + "grad_norm": 97.10274345850128, + "learning_rate": 1.9090565868141093e-07, + "logits/chosen": -1.415880560874939, + "logits/rejected": -1.4168412685394287, + "logps/chosen": -200.85858154296875, + "logps/rejected": -214.74472045898438, + "loss": 0.6936, + "rewards/accuracies": 0.5, + "rewards/chosen": -0.0973915159702301, + "rewards/margins": -0.0053062885999679565, + "rewards/rejected": -0.09208524972200394, + "step": 326 + }, + { + "epoch": 0.3740289359275889, + "grad_norm": 79.53686855879208, + "learning_rate": 1.9074652506433367e-07, + "logits/chosen": -1.2665050029754639, + "logits/rejected": -1.2990330457687378, + "logps/chosen": -188.47377014160156, + "logps/rejected": -196.28854370117188, + "loss": 0.6894, + "rewards/accuracies": 0.53125, + "rewards/chosen": -0.2655767500400543, + "rewards/margins": 0.0786442756652832, + "rewards/rejected": -0.3442210257053375, + "step": 328 + }, + { + "epoch": 0.3763096001710498, + "grad_norm": 92.25716890573669, + "learning_rate": 1.9058607873623696e-07, + "logits/chosen": -1.2212181091308594, + "logits/rejected": -1.2750239372253418, + "logps/chosen": -136.28663635253906, + "logps/rejected": -182.32699584960938, + "loss": 0.6309, + "rewards/accuracies": 0.5, + "rewards/chosen": -0.14970597624778748, + "rewards/margins": 0.13012784719467163, + "rewards/rejected": -0.2798338234424591, + "step": 330 + }, + { + "epoch": 0.3785902644145107, + "grad_norm": 96.38565738306772, + "learning_rate": 1.9042432201808753e-07, + "logits/chosen": -1.3811233043670654, + "logits/rejected": -1.3999402523040771, + "logps/chosen": -153.0084686279297, + "logps/rejected": -174.6513671875, + "loss": 0.6334, + "rewards/accuracies": 0.5625, + "rewards/chosen": -0.06558946520090103, + "rewards/margins": 0.20774412155151367, + "rewards/rejected": -0.2733335494995117, + "step": 332 + }, + { + "epoch": 0.3808709286579716, + "grad_norm": 101.76003883417016, + "learning_rate": 1.9026125724980777e-07, + "logits/chosen": -1.3316706418991089, + "logits/rejected": -1.2979425191879272, + "logps/chosen": -172.97792053222656, + "logps/rejected": -187.61904907226562, + "loss": 0.635, + "rewards/accuracies": 0.65625, + "rewards/chosen": 0.022555537521839142, + "rewards/margins": 0.31808096170425415, + "rewards/rejected": -0.2955254316329956, + "step": 334 + }, + { + "epoch": 0.3831515929014325, + "grad_norm": 89.57559954819274, + "learning_rate": 1.900968867902419e-07, + "logits/chosen": -1.392217993736267, + "logits/rejected": -1.4426984786987305, + "logps/chosen": -171.40127563476562, + "logps/rejected": -171.21238708496094, + "loss": 0.6236, + "rewards/accuracies": 0.65625, + "rewards/chosen": 0.11336915194988251, + "rewards/margins": 0.3153633177280426, + "rewards/rejected": -0.2019941657781601, + "step": 336 + }, + { + "epoch": 0.38543225714489343, + "grad_norm": 95.74770834081339, + "learning_rate": 1.899312130171219e-07, + "logits/chosen": -1.410531759262085, + "logits/rejected": -1.4363325834274292, + "logps/chosen": -176.63296508789062, + "logps/rejected": -194.51480102539062, + "loss": 0.6307, + "rewards/accuracies": 0.65625, + "rewards/chosen": -0.0749322921037674, + "rewards/margins": 0.2967851161956787, + "rewards/rejected": -0.3717173933982849, + "step": 338 + }, + { + "epoch": 0.38771292138835434, + "grad_norm": 94.33644084218925, + "learning_rate": 1.897642383270331e-07, + "logits/chosen": -1.4411017894744873, + "logits/rejected": -1.5141332149505615, + "logps/chosen": -220.8849639892578, + "logps/rejected": -249.29640197753906, + "loss": 0.6396, + "rewards/accuracies": 0.625, + "rewards/chosen": -0.1280645728111267, + "rewards/margins": 0.4015355110168457, + "rewards/rejected": -0.5296000838279724, + "step": 340 + }, + { + "epoch": 0.38999358563181524, + "grad_norm": 93.41769187789713, + "learning_rate": 1.8959596513537937e-07, + "logits/chosen": -1.4379459619522095, + "logits/rejected": -1.463719367980957, + "logps/chosen": -136.88717651367188, + "logps/rejected": -146.65672302246094, + "loss": 0.6624, + "rewards/accuracies": 0.59375, + "rewards/chosen": 0.024333830922842026, + "rewards/margins": 0.15197159349918365, + "rewards/rejected": -0.12763777375221252, + "step": 342 + }, + { + "epoch": 0.39227424987527615, + "grad_norm": 110.48154958430904, + "learning_rate": 1.894263958763485e-07, + "logits/chosen": -1.477798581123352, + "logits/rejected": -1.4197784662246704, + "logps/chosen": -192.24853515625, + "logps/rejected": -182.48162841796875, + "loss": 0.6866, + "rewards/accuracies": 0.625, + "rewards/chosen": 0.029505489394068718, + "rewards/margins": 0.263489693403244, + "rewards/rejected": -0.23398420214653015, + "step": 344 + }, + { + "epoch": 0.39455491411873705, + "grad_norm": 120.46352804291743, + "learning_rate": 1.892555330028766e-07, + "logits/chosen": -1.2521153688430786, + "logits/rejected": -1.2826614379882812, + "logps/chosen": -217.5478515625, + "logps/rejected": -273.1899719238281, + "loss": 0.6418, + "rewards/accuracies": 0.65625, + "rewards/chosen": -0.2787259817123413, + "rewards/margins": 0.3908158242702484, + "rewards/rejected": -0.6695418357849121, + "step": 346 + }, + { + "epoch": 0.396835578362198, + "grad_norm": 94.24036369589392, + "learning_rate": 1.8908337898661285e-07, + "logits/chosen": -1.3746845722198486, + "logits/rejected": -1.46570885181427, + "logps/chosen": -179.17138671875, + "logps/rejected": -217.95236206054688, + "loss": 0.6201, + "rewards/accuracies": 0.75, + "rewards/chosen": -0.12253744155168533, + "rewards/margins": 0.39310359954833984, + "rewards/rejected": -0.5156410336494446, + "step": 348 + }, + { + "epoch": 0.3991162426056589, + "grad_norm": 88.9093668290791, + "learning_rate": 1.889099363178838e-07, + "logits/chosen": -1.445816993713379, + "logits/rejected": -1.4381108283996582, + "logps/chosen": -132.04539489746094, + "logps/rejected": -139.59835815429688, + "loss": 0.6502, + "rewards/accuracies": 0.625, + "rewards/chosen": -0.10454928129911423, + "rewards/margins": 0.15687233209609985, + "rewards/rejected": -0.2614216208457947, + "step": 350 + }, + { + "epoch": 0.40139690684911983, + "grad_norm": 118.96237042903994, + "learning_rate": 1.8873520750565714e-07, + "logits/chosen": -1.3293050527572632, + "logits/rejected": -1.3804844617843628, + "logps/chosen": -172.6136474609375, + "logps/rejected": -184.08212280273438, + "loss": 0.6857, + "rewards/accuracies": 0.53125, + "rewards/chosen": -0.09095679968595505, + "rewards/margins": 0.1739913374185562, + "rewards/rejected": -0.26494812965393066, + "step": 352 + }, + { + "epoch": 0.40367757109258073, + "grad_norm": 123.21030025798191, + "learning_rate": 1.8855919507750556e-07, + "logits/chosen": -1.299286127090454, + "logits/rejected": -1.2867302894592285, + "logps/chosen": -228.6288604736328, + "logps/rejected": -253.30572509765625, + "loss": 0.6819, + "rewards/accuracies": 0.53125, + "rewards/chosen": -0.27850618958473206, + "rewards/margins": 0.12168295681476593, + "rewards/rejected": -0.40018919110298157, + "step": 354 + }, + { + "epoch": 0.40595823533604164, + "grad_norm": 93.75103343413197, + "learning_rate": 1.883819015795701e-07, + "logits/chosen": -1.3721405267715454, + "logits/rejected": -1.4811532497406006, + "logps/chosen": -158.45986938476562, + "logps/rejected": -191.99571228027344, + "loss": 0.6742, + "rewards/accuracies": 0.5625, + "rewards/chosen": 0.06369534879922867, + "rewards/margins": 0.13870595395565033, + "rewards/rejected": -0.07501061260700226, + "step": 356 + }, + { + "epoch": 0.40823889957950255, + "grad_norm": 102.06521296332116, + "learning_rate": 1.8820332957652343e-07, + "logits/chosen": -1.3636136054992676, + "logits/rejected": -1.4187768697738647, + "logps/chosen": -145.48468017578125, + "logps/rejected": -148.94664001464844, + "loss": 0.6414, + "rewards/accuracies": 0.59375, + "rewards/chosen": 0.0663696825504303, + "rewards/margins": 0.29655539989471436, + "rewards/rejected": -0.23018570244312286, + "step": 358 + }, + { + "epoch": 0.41051956382296345, + "grad_norm": 80.21120922967526, + "learning_rate": 1.8802348165153257e-07, + "logits/chosen": -1.0907145738601685, + "logits/rejected": -1.2451905012130737, + "logps/chosen": -140.98895263671875, + "logps/rejected": -169.56390380859375, + "loss": 0.6267, + "rewards/accuracies": 0.625, + "rewards/chosen": 0.1324671506881714, + "rewards/margins": 0.2304615080356598, + "rewards/rejected": -0.0979943573474884, + "step": 360 + }, + { + "epoch": 0.41280022806642436, + "grad_norm": 90.74971301077287, + "learning_rate": 1.8784236040622173e-07, + "logits/chosen": -1.4046244621276855, + "logits/rejected": -1.5294733047485352, + "logps/chosen": -220.828125, + "logps/rejected": -252.51620483398438, + "loss": 0.6485, + "rewards/accuracies": 0.5, + "rewards/chosen": -0.029995199292898178, + "rewards/margins": 0.10216841101646423, + "rewards/rejected": -0.1321636140346527, + "step": 362 + }, + { + "epoch": 0.41508089230988526, + "grad_norm": 116.10619062475209, + "learning_rate": 1.8765996846063453e-07, + "logits/chosen": -1.3053789138793945, + "logits/rejected": -1.331100344657898, + "logps/chosen": -145.552978515625, + "logps/rejected": -175.96580505371094, + "loss": 0.7355, + "rewards/accuracies": 0.46875, + "rewards/chosen": -0.08064378052949905, + "rewards/margins": 0.18256865441799164, + "rewards/rejected": -0.2632124125957489, + "step": 364 + }, + { + "epoch": 0.41736155655334617, + "grad_norm": 85.92321632812764, + "learning_rate": 1.8747630845319612e-07, + "logits/chosen": -1.3430732488632202, + "logits/rejected": -1.321002721786499, + "logps/chosen": -149.8716278076172, + "logps/rejected": -168.82972717285156, + "loss": 0.6414, + "rewards/accuracies": 0.6875, + "rewards/chosen": 0.08440528810024261, + "rewards/margins": 0.3168274164199829, + "rewards/rejected": -0.23242215812206268, + "step": 366 + }, + { + "epoch": 0.4196422207968071, + "grad_norm": 89.28728181459165, + "learning_rate": 1.8729138304067512e-07, + "logits/chosen": -1.464853286743164, + "logits/rejected": -1.478898286819458, + "logps/chosen": -157.5703887939453, + "logps/rejected": -171.20423889160156, + "loss": 0.6186, + "rewards/accuracies": 0.625, + "rewards/chosen": 0.0049234069883823395, + "rewards/margins": 0.2588205635547638, + "rewards/rejected": -0.25389716029167175, + "step": 368 + }, + { + "epoch": 0.421922885040268, + "grad_norm": 86.83502819065235, + "learning_rate": 1.8710519489814503e-07, + "logits/chosen": -1.2088496685028076, + "logits/rejected": -1.238593578338623, + "logps/chosen": -156.05690002441406, + "logps/rejected": -176.14474487304688, + "loss": 0.7105, + "rewards/accuracies": 0.5625, + "rewards/chosen": -0.01236521452665329, + "rewards/margins": 0.17280086874961853, + "rewards/rejected": -0.18516604602336884, + "step": 370 + }, + { + "epoch": 0.4242035492837289, + "grad_norm": 101.40852416620743, + "learning_rate": 1.869177467189456e-07, + "logits/chosen": -1.4426904916763306, + "logits/rejected": -1.409334659576416, + "logps/chosen": -182.2246551513672, + "logps/rejected": -194.63980102539062, + "loss": 0.6937, + "rewards/accuracies": 0.5625, + "rewards/chosen": -0.13504472374916077, + "rewards/margins": 0.12622104585170746, + "rewards/rejected": -0.2612657845020294, + "step": 372 + }, + { + "epoch": 0.4264842135271898, + "grad_norm": 78.92812775165392, + "learning_rate": 1.8672904121464402e-07, + "logits/chosen": -1.3653035163879395, + "logits/rejected": -1.3275575637817383, + "logps/chosen": -180.02243041992188, + "logps/rejected": -195.43106079101562, + "loss": 0.6208, + "rewards/accuracies": 0.6875, + "rewards/chosen": -0.13813042640686035, + "rewards/margins": 0.25349733233451843, + "rewards/rejected": -0.3916277587413788, + "step": 374 + }, + { + "epoch": 0.4287648777706507, + "grad_norm": 85.95171696022348, + "learning_rate": 1.8653908111499531e-07, + "logits/chosen": -1.4379856586456299, + "logits/rejected": -1.5062627792358398, + "logps/chosen": -123.3829345703125, + "logps/rejected": -155.55897521972656, + "loss": 0.6752, + "rewards/accuracies": 0.59375, + "rewards/chosen": -0.02849392406642437, + "rewards/margins": 0.1816612333059311, + "rewards/rejected": -0.2101551592350006, + "step": 376 + }, + { + "epoch": 0.4310455420141116, + "grad_norm": 74.55363107506629, + "learning_rate": 1.8634786916790331e-07, + "logits/chosen": -1.306863784790039, + "logits/rejected": -1.4182804822921753, + "logps/chosen": -185.36831665039062, + "logps/rejected": -205.2194366455078, + "loss": 0.602, + "rewards/accuracies": 0.65625, + "rewards/chosen": 0.23371680080890656, + "rewards/margins": 0.4712272882461548, + "rewards/rejected": -0.23751050233840942, + "step": 378 + }, + { + "epoch": 0.4333262062575725, + "grad_norm": 98.09376209675088, + "learning_rate": 1.861554081393806e-07, + "logits/chosen": -1.298234462738037, + "logits/rejected": -1.320231556892395, + "logps/chosen": -170.1300506591797, + "logps/rejected": -201.2412567138672, + "loss": 0.6166, + "rewards/accuracies": 0.625, + "rewards/chosen": 0.11016674339771271, + "rewards/margins": 0.3684804141521454, + "rewards/rejected": -0.25831368565559387, + "step": 380 + }, + { + "epoch": 0.4356068705010334, + "grad_norm": 95.75031333883096, + "learning_rate": 1.8596170081350855e-07, + "logits/chosen": -1.4295341968536377, + "logits/rejected": -1.4480311870574951, + "logps/chosen": -158.36959838867188, + "logps/rejected": -185.86700439453125, + "loss": 0.6506, + "rewards/accuracies": 0.5625, + "rewards/chosen": 0.11881276965141296, + "rewards/margins": 0.2408534288406372, + "rewards/rejected": -0.12204068899154663, + "step": 382 + }, + { + "epoch": 0.4378875347444943, + "grad_norm": 89.97741935605185, + "learning_rate": 1.8576674999239713e-07, + "logits/chosen": -1.2293694019317627, + "logits/rejected": -1.319061040878296, + "logps/chosen": -168.48960876464844, + "logps/rejected": -213.40200805664062, + "loss": 0.6379, + "rewards/accuracies": 0.71875, + "rewards/chosen": 0.07465538382530212, + "rewards/margins": 0.39287707209587097, + "rewards/rejected": -0.31822171807289124, + "step": 384 + }, + { + "epoch": 0.44016819898795523, + "grad_norm": 84.57348322808303, + "learning_rate": 1.8557055849614428e-07, + "logits/chosen": -1.2857288122177124, + "logits/rejected": -1.317692518234253, + "logps/chosen": -195.7303009033203, + "logps/rejected": -239.7130126953125, + "loss": 0.5716, + "rewards/accuracies": 0.65625, + "rewards/chosen": -0.07659934461116791, + "rewards/margins": 0.5603799819946289, + "rewards/rejected": -0.636979341506958, + "step": 386 + }, + { + "epoch": 0.44244886323141613, + "grad_norm": 85.15633806373397, + "learning_rate": 1.8537312916279523e-07, + "logits/chosen": -1.3817092180252075, + "logits/rejected": -1.501309871673584, + "logps/chosen": -146.41033935546875, + "logps/rejected": -171.60218811035156, + "loss": 0.6673, + "rewards/accuracies": 0.59375, + "rewards/chosen": -0.07926470041275024, + "rewards/margins": 0.26870349049568176, + "rewards/rejected": -0.347968190908432, + "step": 388 + }, + { + "epoch": 0.44472952747487704, + "grad_norm": 89.75323963715459, + "learning_rate": 1.8517446484830136e-07, + "logits/chosen": -1.4567725658416748, + "logits/rejected": -1.5358428955078125, + "logps/chosen": -172.92111206054688, + "logps/rejected": -192.05184936523438, + "loss": 0.6553, + "rewards/accuracies": 0.65625, + "rewards/chosen": -0.1497884839773178, + "rewards/margins": 0.06301924586296082, + "rewards/rejected": -0.21280772984027863, + "step": 390 + }, + { + "epoch": 0.44701019171833795, + "grad_norm": 105.51876340320463, + "learning_rate": 1.8497456842647878e-07, + "logits/chosen": -1.2944555282592773, + "logits/rejected": -1.3529748916625977, + "logps/chosen": -157.77642822265625, + "logps/rejected": -176.93093872070312, + "loss": 0.682, + "rewards/accuracies": 0.5625, + "rewards/chosen": -0.07403340190649033, + "rewards/margins": 0.10720521211624146, + "rewards/rejected": -0.18123860657215118, + "step": 392 + }, + { + "epoch": 0.44929085596179885, + "grad_norm": 85.63092920252892, + "learning_rate": 1.8477344278896706e-07, + "logits/chosen": -1.4767736196517944, + "logits/rejected": -1.5019617080688477, + "logps/chosen": -173.24639892578125, + "logps/rejected": -193.271240234375, + "loss": 0.6935, + "rewards/accuracies": 0.65625, + "rewards/chosen": 0.13190394639968872, + "rewards/margins": 0.48726946115493774, + "rewards/rejected": -0.355365514755249, + "step": 394 + }, + { + "epoch": 0.45157152020525976, + "grad_norm": 100.4356452029139, + "learning_rate": 1.8457109084518718e-07, + "logits/chosen": -1.4190278053283691, + "logits/rejected": -1.3654460906982422, + "logps/chosen": -227.8990020751953, + "logps/rejected": -215.39312744140625, + "loss": 0.6697, + "rewards/accuracies": 0.6875, + "rewards/chosen": -0.15163551270961761, + "rewards/margins": 0.19186191260814667, + "rewards/rejected": -0.3434974253177643, + "step": 396 + }, + { + "epoch": 0.45385218444872066, + "grad_norm": 89.70522326638444, + "learning_rate": 1.8436751552229938e-07, + "logits/chosen": -1.3089743852615356, + "logits/rejected": -1.340443730354309, + "logps/chosen": -145.11398315429688, + "logps/rejected": -172.60635375976562, + "loss": 0.6571, + "rewards/accuracies": 0.75, + "rewards/chosen": -0.20591212809085846, + "rewards/margins": 0.34871169924736023, + "rewards/rejected": -0.5546237230300903, + "step": 398 + }, + { + "epoch": 0.4561328486921816, + "grad_norm": 84.89862446594006, + "learning_rate": 1.84162719765161e-07, + "logits/chosen": -1.2627918720245361, + "logits/rejected": -1.2893693447113037, + "logps/chosen": -130.3240966796875, + "logps/rejected": -160.49227905273438, + "loss": 0.6402, + "rewards/accuracies": 0.78125, + "rewards/chosen": 0.0029626190662384033, + "rewards/margins": 0.4357815384864807, + "rewards/rejected": -0.4328188896179199, + "step": 400 + }, + { + "epoch": 0.4561328486921816, + "eval_logits/chosen": -1.4406086206436157, + "eval_logits/rejected": -1.4253712892532349, + "eval_logps/chosen": -129.20408630371094, + "eval_logps/rejected": -132.1772003173828, + "eval_loss": 0.659533679485321, + "eval_rewards/accuracies": 0.6399999856948853, + "eval_rewards/chosen": 0.06938672810792923, + "eval_rewards/margins": 0.08083352446556091, + "eval_rewards/rejected": -0.011446798220276833, + "eval_runtime": 21.3693, + "eval_samples_per_second": 4.68, + "eval_steps_per_second": 1.17, + "step": 400 + }, + { + "epoch": 0.45841351293564253, + "grad_norm": 86.51322437777803, + "learning_rate": 1.839567065362838e-07, + "logits/chosen": -1.368638038635254, + "logits/rejected": -1.3851830959320068, + "logps/chosen": -169.3848876953125, + "logps/rejected": -175.4376678466797, + "loss": 0.635, + "rewards/accuracies": 0.625, + "rewards/chosen": -0.0501415841281414, + "rewards/margins": 0.1427716761827469, + "rewards/rejected": -0.1929132342338562, + "step": 402 + }, + { + "epoch": 0.46069417717910344, + "grad_norm": 101.82366291792587, + "learning_rate": 1.8374947881579112e-07, + "logits/chosen": -1.201027512550354, + "logits/rejected": -1.2725883722305298, + "logps/chosen": -191.0504150390625, + "logps/rejected": -233.1793212890625, + "loss": 0.6195, + "rewards/accuracies": 0.625, + "rewards/chosen": -0.2396811544895172, + "rewards/margins": 0.2062177062034607, + "rewards/rejected": -0.4458988606929779, + "step": 404 + }, + { + "epoch": 0.46297484142256434, + "grad_norm": 108.6063244513385, + "learning_rate": 1.8354103960137473e-07, + "logits/chosen": -1.4036638736724854, + "logits/rejected": -1.4488118886947632, + "logps/chosen": -230.92767333984375, + "logps/rejected": -239.439208984375, + "loss": 0.6795, + "rewards/accuracies": 0.6875, + "rewards/chosen": -0.32087260484695435, + "rewards/margins": 0.10698382556438446, + "rewards/rejected": -0.4278564453125, + "step": 406 + }, + { + "epoch": 0.46525550566602525, + "grad_norm": 86.59174830548812, + "learning_rate": 1.833313919082515e-07, + "logits/chosen": -1.1800165176391602, + "logits/rejected": -1.3113179206848145, + "logps/chosen": -161.62985229492188, + "logps/rejected": -202.1416473388672, + "loss": 0.6378, + "rewards/accuracies": 0.65625, + "rewards/chosen": -0.14624057710170746, + "rewards/margins": 0.3915923237800598, + "rewards/rejected": -0.5378329157829285, + "step": 408 + }, + { + "epoch": 0.46753616990948615, + "grad_norm": 92.36447587751275, + "learning_rate": 1.8312053876911977e-07, + "logits/chosen": -1.360296368598938, + "logits/rejected": -1.3766343593597412, + "logps/chosen": -203.7003173828125, + "logps/rejected": -233.6885223388672, + "loss": 0.6525, + "rewards/accuracies": 0.5, + "rewards/chosen": -0.13416332006454468, + "rewards/margins": 0.36563166975975037, + "rewards/rejected": -0.49979501962661743, + "step": 410 + }, + { + "epoch": 0.46981683415294706, + "grad_norm": 79.86005120772512, + "learning_rate": 1.8290848323411553e-07, + "logits/chosen": -1.4597082138061523, + "logits/rejected": -1.4648618698120117, + "logps/chosen": -151.07449340820312, + "logps/rejected": -151.00494384765625, + "loss": 0.7101, + "rewards/accuracies": 0.65625, + "rewards/chosen": 0.05570954829454422, + "rewards/margins": 0.08499263972043991, + "rewards/rejected": -0.02928308956325054, + "step": 412 + }, + { + "epoch": 0.47209749839640797, + "grad_norm": 70.72545574881394, + "learning_rate": 1.8269522837076817e-07, + "logits/chosen": -1.4345017671585083, + "logits/rejected": -1.498835802078247, + "logps/chosen": -180.71951293945312, + "logps/rejected": -195.18490600585938, + "loss": 0.6126, + "rewards/accuracies": 0.5625, + "rewards/chosen": -0.060558006167411804, + "rewards/margins": 0.23918625712394714, + "rewards/rejected": -0.29974424839019775, + "step": 414 + }, + { + "epoch": 0.47437816263986887, + "grad_norm": 91.45944584413142, + "learning_rate": 1.8248077726395631e-07, + "logits/chosen": -1.381393313407898, + "logits/rejected": -1.3962711095809937, + "logps/chosen": -139.09548950195312, + "logps/rejected": -154.32943725585938, + "loss": 0.644, + "rewards/accuracies": 0.78125, + "rewards/chosen": 0.08523928374052048, + "rewards/margins": 0.4487174153327942, + "rewards/rejected": -0.3634781539440155, + "step": 416 + }, + { + "epoch": 0.4766588268833298, + "grad_norm": 118.15308317787861, + "learning_rate": 1.8226513301586297e-07, + "logits/chosen": -1.1643140316009521, + "logits/rejected": -1.1523492336273193, + "logps/chosen": -181.3363494873047, + "logps/rejected": -208.14474487304688, + "loss": 0.6631, + "rewards/accuracies": 0.5, + "rewards/chosen": -0.18151137232780457, + "rewards/margins": 0.11719075590372086, + "rewards/rejected": -0.2987021207809448, + "step": 418 + }, + { + "epoch": 0.4789394911267907, + "grad_norm": 87.18607024470172, + "learning_rate": 1.820482987459308e-07, + "logits/chosen": -1.2820664644241333, + "logits/rejected": -1.340864658355713, + "logps/chosen": -152.3217315673828, + "logps/rejected": -166.40199279785156, + "loss": 0.6306, + "rewards/accuracies": 0.5, + "rewards/chosen": -0.20804628729820251, + "rewards/margins": 0.08607995510101318, + "rewards/rejected": -0.2941262423992157, + "step": 420 + }, + { + "epoch": 0.4812201553702516, + "grad_norm": 101.2197264458815, + "learning_rate": 1.818302775908169e-07, + "logits/chosen": -1.3642746210098267, + "logits/rejected": -1.4530210494995117, + "logps/chosen": -177.27783203125, + "logps/rejected": -209.4615478515625, + "loss": 0.6002, + "rewards/accuracies": 0.65625, + "rewards/chosen": -0.06609566509723663, + "rewards/margins": 0.24875135719776154, + "rewards/rejected": -0.31484702229499817, + "step": 422 + }, + { + "epoch": 0.4835008196137125, + "grad_norm": 102.24009472626234, + "learning_rate": 1.8161107270434757e-07, + "logits/chosen": -1.322772741317749, + "logits/rejected": -1.3549699783325195, + "logps/chosen": -175.10890197753906, + "logps/rejected": -189.1114959716797, + "loss": 0.7052, + "rewards/accuracies": 0.53125, + "rewards/chosen": -0.2034468799829483, + "rewards/margins": -0.003699667751789093, + "rewards/rejected": -0.199747234582901, + "step": 424 + }, + { + "epoch": 0.4857814838571734, + "grad_norm": 88.40359207083654, + "learning_rate": 1.8139068725747251e-07, + "logits/chosen": -1.2980220317840576, + "logits/rejected": -1.328129768371582, + "logps/chosen": -127.48295593261719, + "logps/rejected": -145.63137817382812, + "loss": 0.6353, + "rewards/accuracies": 0.5625, + "rewards/chosen": -0.17486554384231567, + "rewards/margins": 0.16773542761802673, + "rewards/rejected": -0.3426010012626648, + "step": 426 + }, + { + "epoch": 0.4880621481006343, + "grad_norm": 85.01546232524993, + "learning_rate": 1.811691244382191e-07, + "logits/chosen": -1.240834355354309, + "logits/rejected": -1.315454125404358, + "logps/chosen": -171.84417724609375, + "logps/rejected": -196.60919189453125, + "loss": 0.6332, + "rewards/accuracies": 0.625, + "rewards/chosen": -0.043825387954711914, + "rewards/margins": 0.35121482610702515, + "rewards/rejected": -0.39504021406173706, + "step": 428 + }, + { + "epoch": 0.4903428123440952, + "grad_norm": 92.50180646153169, + "learning_rate": 1.8094638745164619e-07, + "logits/chosen": -1.368233561515808, + "logits/rejected": -1.4737908840179443, + "logps/chosen": -231.7982177734375, + "logps/rejected": -257.540771484375, + "loss": 0.6393, + "rewards/accuracies": 0.5, + "rewards/chosen": -0.41820961236953735, + "rewards/margins": 0.03227778524160385, + "rewards/rejected": -0.4504873752593994, + "step": 430 + }, + { + "epoch": 0.4926234765875561, + "grad_norm": 93.17182791936958, + "learning_rate": 1.8072247951979782e-07, + "logits/chosen": -1.218642234802246, + "logits/rejected": -1.3522933721542358, + "logps/chosen": -161.9176025390625, + "logps/rejected": -191.59205627441406, + "loss": 0.6532, + "rewards/accuracies": 0.46875, + "rewards/chosen": -0.2410704791545868, + "rewards/margins": 0.09214940667152405, + "rewards/rejected": -0.33321988582611084, + "step": 432 + }, + { + "epoch": 0.494904140831017, + "grad_norm": 85.22671128905367, + "learning_rate": 1.8049740388165646e-07, + "logits/chosen": -1.3610548973083496, + "logits/rejected": -1.369373083114624, + "logps/chosen": -170.4285888671875, + "logps/rejected": -203.44435119628906, + "loss": 0.6788, + "rewards/accuracies": 0.53125, + "rewards/chosen": -0.12429852783679962, + "rewards/margins": 0.06445370614528656, + "rewards/rejected": -0.18875223398208618, + "step": 434 + }, + { + "epoch": 0.49718480507447793, + "grad_norm": 106.2899671913355, + "learning_rate": 1.8027116379309635e-07, + "logits/chosen": -1.3293408155441284, + "logits/rejected": -1.3227120637893677, + "logps/chosen": -189.62835693359375, + "logps/rejected": -196.80776977539062, + "loss": 0.6658, + "rewards/accuracies": 0.4375, + "rewards/chosen": -0.36924922466278076, + "rewards/margins": 0.04584325850009918, + "rewards/rejected": -0.41509246826171875, + "step": 436 + }, + { + "epoch": 0.49946546931793884, + "grad_norm": 97.4627918918024, + "learning_rate": 1.8004376252683629e-07, + "logits/chosen": -1.3211669921875, + "logits/rejected": -1.318382740020752, + "logps/chosen": -170.44007873535156, + "logps/rejected": -177.15567016601562, + "loss": 0.652, + "rewards/accuracies": 0.65625, + "rewards/chosen": -0.0650164932012558, + "rewards/margins": 0.2474796026945114, + "rewards/rejected": -0.3124960958957672, + "step": 438 + }, + { + "epoch": 0.5017461335613997, + "grad_norm": 94.35984756124336, + "learning_rate": 1.7981520337239229e-07, + "logits/chosen": -1.4195128679275513, + "logits/rejected": -1.517584204673767, + "logps/chosen": -124.39069366455078, + "logps/rejected": -148.39462280273438, + "loss": 0.6071, + "rewards/accuracies": 0.65625, + "rewards/chosen": 0.07836849987506866, + "rewards/margins": 0.2909252941608429, + "rewards/rejected": -0.21255679428577423, + "step": 440 + }, + { + "epoch": 0.5040267978048607, + "grad_norm": 105.60849785169918, + "learning_rate": 1.7958548963603e-07, + "logits/chosen": -1.3688621520996094, + "logits/rejected": -1.3791152238845825, + "logps/chosen": -193.5460968017578, + "logps/rejected": -198.70162963867188, + "loss": 0.6155, + "rewards/accuracies": 0.8125, + "rewards/chosen": 0.006075154058635235, + "rewards/margins": 0.4186677634716034, + "rewards/rejected": -0.41259267926216125, + "step": 442 + }, + { + "epoch": 0.5063074620483216, + "grad_norm": 105.19505214300226, + "learning_rate": 1.7935462464071694e-07, + "logits/chosen": -1.4063529968261719, + "logits/rejected": -1.4609131813049316, + "logps/chosen": -136.6985321044922, + "logps/rejected": -142.0517120361328, + "loss": 0.6638, + "rewards/accuracies": 0.71875, + "rewards/chosen": -0.021208282560110092, + "rewards/margins": 0.23011060059070587, + "rewards/rejected": -0.25131887197494507, + "step": 444 + }, + { + "epoch": 0.5085881262917825, + "grad_norm": 80.98126219811707, + "learning_rate": 1.7912261172607434e-07, + "logits/chosen": -1.4103684425354004, + "logits/rejected": -1.4659305810928345, + "logps/chosen": -181.66671752929688, + "logps/rejected": -211.0421142578125, + "loss": 0.667, + "rewards/accuracies": 0.6875, + "rewards/chosen": -0.0669156014919281, + "rewards/margins": 0.23970045149326324, + "rewards/rejected": -0.30661600828170776, + "step": 446 + }, + { + "epoch": 0.5108687905352434, + "grad_norm": 86.7187808268374, + "learning_rate": 1.7888945424832892e-07, + "logits/chosen": -1.2415688037872314, + "logits/rejected": -1.2903611660003662, + "logps/chosen": -195.07174682617188, + "logps/rejected": -223.18804931640625, + "loss": 0.6344, + "rewards/accuracies": 0.625, + "rewards/chosen": -0.26416096091270447, + "rewards/margins": 0.23690298199653625, + "rewards/rejected": -0.5010639429092407, + "step": 448 + }, + { + "epoch": 0.5131494547787043, + "grad_norm": 91.51916447980646, + "learning_rate": 1.7865515558026425e-07, + "logits/chosen": -1.2545722723007202, + "logits/rejected": -1.320673942565918, + "logps/chosen": -190.4950714111328, + "logps/rejected": -196.8434600830078, + "loss": 0.645, + "rewards/accuracies": 0.6875, + "rewards/chosen": -0.12865591049194336, + "rewards/margins": 0.2827560007572174, + "rewards/rejected": -0.41141194105148315, + "step": 450 + }, + { + "epoch": 0.5154301190221652, + "grad_norm": 78.95115222899102, + "learning_rate": 1.78419719111172e-07, + "logits/chosen": -1.2161290645599365, + "logits/rejected": -1.2955366373062134, + "logps/chosen": -133.3023223876953, + "logps/rejected": -173.38735961914062, + "loss": 0.6202, + "rewards/accuracies": 0.625, + "rewards/chosen": -0.031168239191174507, + "rewards/margins": 0.3425666093826294, + "rewards/rejected": -0.37373483180999756, + "step": 452 + }, + { + "epoch": 0.5177107832656261, + "grad_norm": 85.03550287122602, + "learning_rate": 1.78183148246803e-07, + "logits/chosen": -1.3149534463882446, + "logits/rejected": -1.3734480142593384, + "logps/chosen": -181.44749450683594, + "logps/rejected": -190.66098022460938, + "loss": 0.6579, + "rewards/accuracies": 0.4375, + "rewards/chosen": -0.02150813303887844, + "rewards/margins": 0.09824002534151077, + "rewards/rejected": -0.11974816769361496, + "step": 454 + }, + { + "epoch": 0.519991447509087, + "grad_norm": 101.82433211173363, + "learning_rate": 1.779454464093177e-07, + "logits/chosen": -1.2327746152877808, + "logits/rejected": -1.2479290962219238, + "logps/chosen": -123.62596130371094, + "logps/rejected": -145.822021484375, + "loss": 0.71, + "rewards/accuracies": 0.65625, + "rewards/chosen": 0.008450210094451904, + "rewards/margins": 0.1642961949110031, + "rewards/rejected": -0.1558459848165512, + "step": 456 + }, + { + "epoch": 0.522272111752548, + "grad_norm": 86.25249319253086, + "learning_rate": 1.7770661703723716e-07, + "logits/chosen": -1.4179208278656006, + "logits/rejected": -1.4704089164733887, + "logps/chosen": -155.395263671875, + "logps/rejected": -241.91619873046875, + "loss": 0.6544, + "rewards/accuracies": 0.625, + "rewards/chosen": -0.0192805714905262, + "rewards/margins": 0.4996611773967743, + "rewards/rejected": -0.5189418792724609, + "step": 458 + }, + { + "epoch": 0.5245527759960088, + "grad_norm": 95.73185772448069, + "learning_rate": 1.7746666358539268e-07, + "logits/chosen": -1.324285864830017, + "logits/rejected": -1.3773202896118164, + "logps/chosen": -138.01214599609375, + "logps/rejected": -187.76605224609375, + "loss": 0.6781, + "rewards/accuracies": 0.5, + "rewards/chosen": -0.11806146800518036, + "rewards/margins": 0.12408170104026794, + "rewards/rejected": -0.2421431541442871, + "step": 460 + }, + { + "epoch": 0.5268334402394698, + "grad_norm": 83.04802100256194, + "learning_rate": 1.7722558952487637e-07, + "logits/chosen": -1.2252494096755981, + "logits/rejected": -1.2402747869491577, + "logps/chosen": -145.02134704589844, + "logps/rejected": -157.798095703125, + "loss": 0.6658, + "rewards/accuracies": 0.59375, + "rewards/chosen": -0.16228270530700684, + "rewards/margins": -0.006807304918766022, + "rewards/rejected": -0.1554754078388214, + "step": 462 + }, + { + "epoch": 0.5291141044829306, + "grad_norm": 106.4784679324614, + "learning_rate": 1.7698339834299062e-07, + "logits/chosen": -1.2572526931762695, + "logits/rejected": -1.2763482332229614, + "logps/chosen": -205.0850067138672, + "logps/rejected": -216.61383056640625, + "loss": 0.6758, + "rewards/accuracies": 0.71875, + "rewards/chosen": -0.10254046320915222, + "rewards/margins": 0.47703590989112854, + "rewards/rejected": -0.5795763731002808, + "step": 464 + }, + { + "epoch": 0.5313947687263916, + "grad_norm": 88.23039919938233, + "learning_rate": 1.7674009354319776e-07, + "logits/chosen": -1.2906126976013184, + "logits/rejected": -1.3812601566314697, + "logps/chosen": -182.90606689453125, + "logps/rejected": -192.94520568847656, + "loss": 0.5967, + "rewards/accuracies": 0.59375, + "rewards/chosen": -0.12573322653770447, + "rewards/margins": 0.15527519583702087, + "rewards/rejected": -0.28100842237472534, + "step": 466 + }, + { + "epoch": 0.5336754329698524, + "grad_norm": 96.24054074341065, + "learning_rate": 1.764956786450694e-07, + "logits/chosen": -1.2798047065734863, + "logits/rejected": -1.3678182363510132, + "logps/chosen": -160.85719299316406, + "logps/rejected": -189.01951599121094, + "loss": 0.6321, + "rewards/accuracies": 0.78125, + "rewards/chosen": -0.01916433498263359, + "rewards/margins": 0.328861802816391, + "rewards/rejected": -0.3480261266231537, + "step": 468 + }, + { + "epoch": 0.5359560972133134, + "grad_norm": 87.51604676838089, + "learning_rate": 1.7625015718423548e-07, + "logits/chosen": -1.3382031917572021, + "logits/rejected": -1.4446802139282227, + "logps/chosen": -152.4910125732422, + "logps/rejected": -192.94979858398438, + "loss": 0.6364, + "rewards/accuracies": 0.65625, + "rewards/chosen": -0.021129202097654343, + "rewards/margins": 0.3612152636051178, + "rewards/rejected": -0.38234445452690125, + "step": 470 + }, + { + "epoch": 0.5382367614567742, + "grad_norm": 118.40842425272356, + "learning_rate": 1.7600353271233312e-07, + "logits/chosen": -1.319710373878479, + "logits/rejected": -1.284549355506897, + "logps/chosen": -152.95933532714844, + "logps/rejected": -156.8072052001953, + "loss": 0.7236, + "rewards/accuracies": 0.53125, + "rewards/chosen": -0.10047149658203125, + "rewards/margins": 0.010637946426868439, + "rewards/rejected": -0.11110943555831909, + "step": 472 + }, + { + "epoch": 0.5405174257002352, + "grad_norm": 91.83509455780727, + "learning_rate": 1.7575580879695522e-07, + "logits/chosen": -1.445102572441101, + "logits/rejected": -1.4954876899719238, + "logps/chosen": -196.539306640625, + "logps/rejected": -211.00616455078125, + "loss": 0.6409, + "rewards/accuracies": 0.625, + "rewards/chosen": -0.04536067694425583, + "rewards/margins": 0.1365879625082016, + "rewards/rejected": -0.18194863200187683, + "step": 474 + }, + { + "epoch": 0.542798089943696, + "grad_norm": 90.74830800218245, + "learning_rate": 1.7550698902159894e-07, + "logits/chosen": -1.3519353866577148, + "logits/rejected": -1.4368261098861694, + "logps/chosen": -228.4932098388672, + "logps/rejected": -261.87109375, + "loss": 0.6373, + "rewards/accuracies": 0.46875, + "rewards/chosen": -0.1739276647567749, + "rewards/margins": 0.011357128620147705, + "rewards/rejected": -0.1852847784757614, + "step": 476 + }, + { + "epoch": 0.545078754187157, + "grad_norm": 114.14083612276099, + "learning_rate": 1.7525707698561384e-07, + "logits/chosen": -1.296964168548584, + "logits/rejected": -1.3777580261230469, + "logps/chosen": -178.81484985351562, + "logps/rejected": -220.63436889648438, + "loss": 0.6515, + "rewards/accuracies": 0.75, + "rewards/chosen": -0.10972025990486145, + "rewards/margins": 0.45839735865592957, + "rewards/rejected": -0.5681176781654358, + "step": 478 + }, + { + "epoch": 0.547359418430618, + "grad_norm": 106.70099483754956, + "learning_rate": 1.750060763041497e-07, + "logits/chosen": -1.3242031335830688, + "logits/rejected": -1.3594613075256348, + "logps/chosen": -139.518310546875, + "logps/rejected": -171.3057403564453, + "loss": 0.6932, + "rewards/accuracies": 0.59375, + "rewards/chosen": -0.09948215633630753, + "rewards/margins": 0.13577090203762054, + "rewards/rejected": -0.23525305092334747, + "step": 480 + }, + { + "epoch": 0.5496400826740788, + "grad_norm": 93.76871490458784, + "learning_rate": 1.7475399060810435e-07, + "logits/chosen": -1.5048158168792725, + "logits/rejected": -1.567376971244812, + "logps/chosen": -175.3671112060547, + "logps/rejected": -195.3438262939453, + "loss": 0.6284, + "rewards/accuracies": 0.625, + "rewards/chosen": 0.08378006517887115, + "rewards/margins": 0.32175135612487793, + "rewards/rejected": -0.23797132074832916, + "step": 482 + }, + { + "epoch": 0.5519207469175398, + "grad_norm": 87.30478432077047, + "learning_rate": 1.7450082354407107e-07, + "logits/chosen": -1.3454846143722534, + "logits/rejected": -1.3264262676239014, + "logps/chosen": -150.41058349609375, + "logps/rejected": -158.28414916992188, + "loss": 0.6392, + "rewards/accuracies": 0.53125, + "rewards/chosen": 0.06661003828048706, + "rewards/margins": 0.05708365514874458, + "rewards/rejected": 0.009526383131742477, + "step": 484 + }, + { + "epoch": 0.5542014111610006, + "grad_norm": 97.70119982177076, + "learning_rate": 1.7424657877428594e-07, + "logits/chosen": -1.4438061714172363, + "logits/rejected": -1.5046687126159668, + "logps/chosen": -210.83221435546875, + "logps/rejected": -220.98056030273438, + "loss": 0.6711, + "rewards/accuracies": 0.59375, + "rewards/chosen": 0.02250128984451294, + "rewards/margins": 0.25783565640449524, + "rewards/rejected": -0.23533441126346588, + "step": 486 + }, + { + "epoch": 0.5564820754044616, + "grad_norm": 92.7470076491593, + "learning_rate": 1.7399125997657475e-07, + "logits/chosen": -1.4578030109405518, + "logits/rejected": -1.5047590732574463, + "logps/chosen": -173.8478546142578, + "logps/rejected": -218.70567321777344, + "loss": 0.6464, + "rewards/accuracies": 0.75, + "rewards/chosen": -0.005423933267593384, + "rewards/margins": 0.22333881258964539, + "rewards/rejected": -0.22876277565956116, + "step": 488 + }, + { + "epoch": 0.5587627396479224, + "grad_norm": 80.43868813641923, + "learning_rate": 1.7373487084429986e-07, + "logits/chosen": -1.4293080568313599, + "logits/rejected": -1.3778735399246216, + "logps/chosen": -137.72915649414062, + "logps/rejected": -144.8123779296875, + "loss": 0.6175, + "rewards/accuracies": 0.65625, + "rewards/chosen": -0.04381662234663963, + "rewards/margins": 0.17386358976364136, + "rewards/rejected": -0.2176802158355713, + "step": 490 + }, + { + "epoch": 0.5610434038913834, + "grad_norm": 88.32928179136898, + "learning_rate": 1.734774150863067e-07, + "logits/chosen": -1.4234795570373535, + "logits/rejected": -1.4454480409622192, + "logps/chosen": -168.4951171875, + "logps/rejected": -183.8735809326172, + "loss": 0.6522, + "rewards/accuracies": 0.625, + "rewards/chosen": 0.008489616215229034, + "rewards/margins": 0.20173315703868866, + "rewards/rejected": -0.19324351847171783, + "step": 492 + }, + { + "epoch": 0.5633240681348443, + "grad_norm": 93.76675043932717, + "learning_rate": 1.732188964268703e-07, + "logits/chosen": -1.2950657606124878, + "logits/rejected": -1.3514856100082397, + "logps/chosen": -174.49368286132812, + "logps/rejected": -198.4128875732422, + "loss": 0.6429, + "rewards/accuracies": 0.5, + "rewards/chosen": -0.17520295083522797, + "rewards/margins": 0.17245244979858398, + "rewards/rejected": -0.34765538573265076, + "step": 494 + }, + { + "epoch": 0.5656047323783052, + "grad_norm": 124.66844580791373, + "learning_rate": 1.7295931860564118e-07, + "logits/chosen": -1.3973047733306885, + "logits/rejected": -1.3848158121109009, + "logps/chosen": -187.01031494140625, + "logps/rejected": -201.88232421875, + "loss": 0.661, + "rewards/accuracies": 0.4375, + "rewards/chosen": -0.4809916317462921, + "rewards/margins": -0.037112899124622345, + "rewards/rejected": -0.443878710269928, + "step": 496 + }, + { + "epoch": 0.5678853966217661, + "grad_norm": 98.27416452810739, + "learning_rate": 1.7269868537759138e-07, + "logits/chosen": -1.4414607286453247, + "logits/rejected": -1.4525971412658691, + "logps/chosen": -186.87220764160156, + "logps/rejected": -201.6339874267578, + "loss": 0.6551, + "rewards/accuracies": 0.75, + "rewards/chosen": 0.048535339534282684, + "rewards/margins": 0.2541671097278595, + "rewards/rejected": -0.20563176274299622, + "step": 498 + }, + { + "epoch": 0.570166060865227, + "grad_norm": 81.20755107084783, + "learning_rate": 1.7243700051296016e-07, + "logits/chosen": -1.4566954374313354, + "logits/rejected": -1.458418369293213, + "logps/chosen": -186.07911682128906, + "logps/rejected": -181.54051208496094, + "loss": 0.6716, + "rewards/accuracies": 0.4375, + "rewards/chosen": -0.08042742311954498, + "rewards/margins": 0.003591485321521759, + "rewards/rejected": -0.08401892334222794, + "step": 500 + }, + { + "epoch": 0.570166060865227, + "eval_logits/chosen": -1.4689228534698486, + "eval_logits/rejected": -1.4550352096557617, + "eval_logps/chosen": -128.87640380859375, + "eval_logps/rejected": -132.28384399414062, + "eval_loss": 0.6351009011268616, + "eval_rewards/accuracies": 0.6399999856948853, + "eval_rewards/chosen": 0.10215496271848679, + "eval_rewards/margins": 0.12426460534334183, + "eval_rewards/rejected": -0.022109635174274445, + "eval_runtime": 20.6735, + "eval_samples_per_second": 4.837, + "eval_steps_per_second": 1.209, + "step": 500 + }, + { + "epoch": 0.5724467251086879, + "grad_norm": 87.15380241514247, + "learning_rate": 1.7217426779719944e-07, + "logits/chosen": -1.461560606956482, + "logits/rejected": -1.4978594779968262, + "logps/chosen": -145.8336639404297, + "logps/rejected": -164.410888671875, + "loss": 0.623, + "rewards/accuracies": 0.65625, + "rewards/chosen": -0.030693676322698593, + "rewards/margins": 0.31371259689331055, + "rewards/rejected": -0.34440624713897705, + "step": 502 + }, + { + "epoch": 0.5747273893521488, + "grad_norm": 109.6734083472702, + "learning_rate": 1.71910491030919e-07, + "logits/chosen": -1.4200553894042969, + "logits/rejected": -1.4383351802825928, + "logps/chosen": -185.2432861328125, + "logps/rejected": -198.3214569091797, + "loss": 0.6592, + "rewards/accuracies": 0.5625, + "rewards/chosen": -0.09178508818149567, + "rewards/margins": 0.1750696301460266, + "rewards/rejected": -0.2668547034263611, + "step": 504 + }, + { + "epoch": 0.5770080535956097, + "grad_norm": 95.33984559143799, + "learning_rate": 1.716456740298315e-07, + "logits/chosen": -1.356040120124817, + "logits/rejected": -1.3635413646697998, + "logps/chosen": -221.06320190429688, + "logps/rejected": -246.34091186523438, + "loss": 0.6466, + "rewards/accuracies": 0.59375, + "rewards/chosen": -0.224991112947464, + "rewards/margins": 0.12949411571025848, + "rewards/rejected": -0.35448524355888367, + "step": 506 + }, + { + "epoch": 0.5792887178390707, + "grad_norm": 104.51344570968739, + "learning_rate": 1.7137982062469739e-07, + "logits/chosen": -1.199120044708252, + "logits/rejected": -1.2254160642623901, + "logps/chosen": -182.35678100585938, + "logps/rejected": -205.9598388671875, + "loss": 0.6302, + "rewards/accuracies": 0.6875, + "rewards/chosen": -0.01045935694128275, + "rewards/margins": 0.4567835330963135, + "rewards/rejected": -0.46724286675453186, + "step": 508 + }, + { + "epoch": 0.5815693820825315, + "grad_norm": 80.61166146639357, + "learning_rate": 1.7111293466126936e-07, + "logits/chosen": -1.4290285110473633, + "logits/rejected": -1.465309977531433, + "logps/chosen": -183.41546630859375, + "logps/rejected": -190.38209533691406, + "loss": 0.6303, + "rewards/accuracies": 0.625, + "rewards/chosen": -0.04882081598043442, + "rewards/margins": 0.2510834336280823, + "rewards/rejected": -0.2999042272567749, + "step": 510 + }, + { + "epoch": 0.5838500463259925, + "grad_norm": 81.58673928063743, + "learning_rate": 1.7084502000023678e-07, + "logits/chosen": -1.3187546730041504, + "logits/rejected": -1.3721165657043457, + "logps/chosen": -166.35317993164062, + "logps/rejected": -205.14459228515625, + "loss": 0.6609, + "rewards/accuracies": 0.6875, + "rewards/chosen": 0.06919605284929276, + "rewards/margins": 0.25993865728378296, + "rewards/rejected": -0.1907426118850708, + "step": 512 + }, + { + "epoch": 0.5861307105694533, + "grad_norm": 107.99959199152349, + "learning_rate": 1.7057608051716987e-07, + "logits/chosen": -1.2958874702453613, + "logits/rejected": -1.3409713506698608, + "logps/chosen": -189.3512420654297, + "logps/rejected": -229.98101806640625, + "loss": 0.6805, + "rewards/accuracies": 0.625, + "rewards/chosen": -0.1412532776594162, + "rewards/margins": 0.1916051059961319, + "rewards/rejected": -0.3328583836555481, + "step": 514 + }, + { + "epoch": 0.5884113748129143, + "grad_norm": 103.31085588981476, + "learning_rate": 1.7030612010246357e-07, + "logits/chosen": -1.4542144536972046, + "logits/rejected": -1.4537440538406372, + "logps/chosen": -187.4440155029297, + "logps/rejected": -197.0413818359375, + "loss": 0.6901, + "rewards/accuracies": 0.5625, + "rewards/chosen": -0.20216982066631317, + "rewards/margins": -0.05732645466923714, + "rewards/rejected": -0.14484335482120514, + "step": 516 + }, + { + "epoch": 0.5906920390563751, + "grad_norm": 93.7004047268541, + "learning_rate": 1.7003514266128128e-07, + "logits/chosen": -1.3705557584762573, + "logits/rejected": -1.4125196933746338, + "logps/chosen": -176.84820556640625, + "logps/rejected": -218.4021453857422, + "loss": 0.688, + "rewards/accuracies": 0.5625, + "rewards/chosen": -0.13926447927951813, + "rewards/margins": 0.1602565199136734, + "rewards/rejected": -0.2995210289955139, + "step": 518 + }, + { + "epoch": 0.5929727032998361, + "grad_norm": 76.7210943847488, + "learning_rate": 1.6976315211349848e-07, + "logits/chosen": -1.402063250541687, + "logits/rejected": -1.5004804134368896, + "logps/chosen": -183.45787048339844, + "logps/rejected": -226.31532287597656, + "loss": 0.6405, + "rewards/accuracies": 0.71875, + "rewards/chosen": 0.08890336751937866, + "rewards/margins": 0.2809264659881592, + "rewards/rejected": -0.19202306866645813, + "step": 520 + }, + { + "epoch": 0.5952533675432969, + "grad_norm": 79.99952128240997, + "learning_rate": 1.694901523936458e-07, + "logits/chosen": -1.3303985595703125, + "logits/rejected": -1.4111953973770142, + "logps/chosen": -161.44390869140625, + "logps/rejected": -201.76593017578125, + "loss": 0.601, + "rewards/accuracies": 0.78125, + "rewards/chosen": 0.12671524286270142, + "rewards/margins": 0.4966307282447815, + "rewards/rejected": -0.3699154853820801, + "step": 522 + }, + { + "epoch": 0.5975340317867579, + "grad_norm": 89.60709297255305, + "learning_rate": 1.6921614745085235e-07, + "logits/chosen": -1.4100148677825928, + "logits/rejected": -1.4378656148910522, + "logps/chosen": -193.54835510253906, + "logps/rejected": -202.9269256591797, + "loss": 0.6182, + "rewards/accuracies": 0.71875, + "rewards/chosen": -0.035577308386564255, + "rewards/margins": 0.3006408214569092, + "rewards/rejected": -0.33621811866760254, + "step": 524 + }, + { + "epoch": 0.5998146960302188, + "grad_norm": 104.27911734761277, + "learning_rate": 1.689411412487885e-07, + "logits/chosen": -1.3611271381378174, + "logits/rejected": -1.3725131750106812, + "logps/chosen": -166.821533203125, + "logps/rejected": -197.65394592285156, + "loss": 0.6274, + "rewards/accuracies": 0.65625, + "rewards/chosen": 0.09458069503307343, + "rewards/margins": 0.400374174118042, + "rewards/rejected": -0.30579352378845215, + "step": 526 + }, + { + "epoch": 0.6020953602736797, + "grad_norm": 77.9183795766502, + "learning_rate": 1.6866513776560841e-07, + "logits/chosen": -1.2477551698684692, + "logits/rejected": -1.2647314071655273, + "logps/chosen": -143.03448486328125, + "logps/rejected": -149.70594787597656, + "loss": 0.6035, + "rewards/accuracies": 0.71875, + "rewards/chosen": 0.14375856518745422, + "rewards/margins": 0.17092323303222656, + "rewards/rejected": -0.02716466784477234, + "step": 528 + }, + { + "epoch": 0.6043760245171406, + "grad_norm": 79.08140854679833, + "learning_rate": 1.6838814099389265e-07, + "logits/chosen": -1.379310965538025, + "logits/rejected": -1.49821138381958, + "logps/chosen": -176.78890991210938, + "logps/rejected": -228.0948028564453, + "loss": 0.6114, + "rewards/accuracies": 0.59375, + "rewards/chosen": -0.18548306822776794, + "rewards/margins": 0.3985109329223633, + "rewards/rejected": -0.5839939117431641, + "step": 530 + }, + { + "epoch": 0.6066566887606015, + "grad_norm": 86.55652744872312, + "learning_rate": 1.6811015494059045e-07, + "logits/chosen": -1.340354323387146, + "logits/rejected": -1.372084617614746, + "logps/chosen": -139.92453002929688, + "logps/rejected": -170.04629516601562, + "loss": 0.5944, + "rewards/accuracies": 0.625, + "rewards/chosen": 0.09279327094554901, + "rewards/margins": 0.36332762241363525, + "rewards/rejected": -0.27053436636924744, + "step": 532 + }, + { + "epoch": 0.6089373530040625, + "grad_norm": 86.16548448796566, + "learning_rate": 1.678311836269616e-07, + "logits/chosen": -1.2494535446166992, + "logits/rejected": -1.3522768020629883, + "logps/chosen": -133.36306762695312, + "logps/rejected": -140.9103240966797, + "loss": 0.626, + "rewards/accuracies": 0.625, + "rewards/chosen": -0.22486190497875214, + "rewards/margins": 0.10757801681756973, + "rewards/rejected": -0.3324398994445801, + "step": 534 + }, + { + "epoch": 0.6112180172475233, + "grad_norm": 78.5036790731728, + "learning_rate": 1.6755123108851842e-07, + "logits/chosen": -1.2597969770431519, + "logits/rejected": -1.3842471837997437, + "logps/chosen": -102.91565704345703, + "logps/rejected": -149.77996826171875, + "loss": 0.6558, + "rewards/accuracies": 0.59375, + "rewards/chosen": 0.049101147800683975, + "rewards/margins": 0.2616952359676361, + "rewards/rejected": -0.21259410679340363, + "step": 536 + }, + { + "epoch": 0.6134986814909843, + "grad_norm": 97.00034287566903, + "learning_rate": 1.6727030137496727e-07, + "logits/chosen": -1.333500862121582, + "logits/rejected": -1.407701015472412, + "logps/chosen": -168.34678649902344, + "logps/rejected": -212.6487579345703, + "loss": 0.6333, + "rewards/accuracies": 0.65625, + "rewards/chosen": -0.23587258160114288, + "rewards/margins": 0.40141719579696655, + "rewards/rejected": -0.637289822101593, + "step": 538 + }, + { + "epoch": 0.6157793457344451, + "grad_norm": 93.14382617410926, + "learning_rate": 1.6698839855015007e-07, + "logits/chosen": -1.3826831579208374, + "logits/rejected": -1.3780534267425537, + "logps/chosen": -197.9063262939453, + "logps/rejected": -201.8406982421875, + "loss": 0.653, + "rewards/accuracies": 0.46875, + "rewards/chosen": -0.19411635398864746, + "rewards/margins": 0.06094612553715706, + "rewards/rejected": -0.2550624907016754, + "step": 540 + }, + { + "epoch": 0.6180600099779061, + "grad_norm": 84.06073173280129, + "learning_rate": 1.6670552669198546e-07, + "logits/chosen": -1.459622859954834, + "logits/rejected": -1.5475270748138428, + "logps/chosen": -149.38450622558594, + "logps/rejected": -174.00827026367188, + "loss": 0.6878, + "rewards/accuracies": 0.625, + "rewards/chosen": -0.22387176752090454, + "rewards/margins": 0.04707152396440506, + "rewards/rejected": -0.2709433138370514, + "step": 542 + }, + { + "epoch": 0.620340674221367, + "grad_norm": 75.86401555746195, + "learning_rate": 1.6642168989240974e-07, + "logits/chosen": -1.3617055416107178, + "logits/rejected": -1.3725388050079346, + "logps/chosen": -177.85269165039062, + "logps/rejected": -204.47271728515625, + "loss": 0.6393, + "rewards/accuracies": 0.75, + "rewards/chosen": -0.2369232326745987, + "rewards/margins": 0.4420955181121826, + "rewards/rejected": -0.6790187358856201, + "step": 544 + }, + { + "epoch": 0.6226213384648279, + "grad_norm": 99.08946862124856, + "learning_rate": 1.6613689225731787e-07, + "logits/chosen": -1.437178611755371, + "logits/rejected": -1.4497863054275513, + "logps/chosen": -152.61231994628906, + "logps/rejected": -163.36196899414062, + "loss": 0.6514, + "rewards/accuracies": 0.5625, + "rewards/chosen": -0.24256393313407898, + "rewards/margins": 0.14335207641124725, + "rewards/rejected": -0.3859160244464874, + "step": 546 + }, + { + "epoch": 0.6249020027082888, + "grad_norm": 82.17381709045306, + "learning_rate": 1.6585113790650386e-07, + "logits/chosen": -1.3418258428573608, + "logits/rejected": -1.357533574104309, + "logps/chosen": -150.5135955810547, + "logps/rejected": -168.92135620117188, + "loss": 0.6404, + "rewards/accuracies": 0.625, + "rewards/chosen": -0.1530638188123703, + "rewards/margins": 0.16317017376422882, + "rewards/rejected": -0.3162340223789215, + "step": 548 + }, + { + "epoch": 0.6271826669517497, + "grad_norm": 98.6032079855375, + "learning_rate": 1.6556443097360133e-07, + "logits/chosen": -1.305914282798767, + "logits/rejected": -1.3782196044921875, + "logps/chosen": -187.4833526611328, + "logps/rejected": -213.8814697265625, + "loss": 0.6415, + "rewards/accuracies": 0.75, + "rewards/chosen": -0.03956926241517067, + "rewards/margins": 0.45714423060417175, + "rewards/rejected": -0.4967134892940521, + "step": 550 + }, + { + "epoch": 0.6294633311952106, + "grad_norm": 89.5606898436777, + "learning_rate": 1.6527677560602363e-07, + "logits/chosen": -1.4830090999603271, + "logits/rejected": -1.4891180992126465, + "logps/chosen": -178.29112243652344, + "logps/rejected": -182.18019104003906, + "loss": 0.658, + "rewards/accuracies": 0.6875, + "rewards/chosen": -0.10583019256591797, + "rewards/margins": 0.16508930921554565, + "rewards/rejected": -0.270919531583786, + "step": 552 + }, + { + "epoch": 0.6317439954386715, + "grad_norm": 94.29213072448552, + "learning_rate": 1.6498817596490388e-07, + "logits/chosen": -1.3679401874542236, + "logits/rejected": -1.397881031036377, + "logps/chosen": -170.22218322753906, + "logps/rejected": -190.22608947753906, + "loss": 0.6632, + "rewards/accuracies": 0.625, + "rewards/chosen": -0.07908223569393158, + "rewards/margins": 0.2489246428012848, + "rewards/rejected": -0.32800689339637756, + "step": 554 + }, + { + "epoch": 0.6340246596821324, + "grad_norm": 97.6468147380264, + "learning_rate": 1.6469863622503475e-07, + "logits/chosen": -1.3809480667114258, + "logits/rejected": -1.3862886428833008, + "logps/chosen": -161.6693878173828, + "logps/rejected": -186.37608337402344, + "loss": 0.6762, + "rewards/accuracies": 0.6875, + "rewards/chosen": -0.08833756297826767, + "rewards/margins": 0.3514275848865509, + "rewards/rejected": -0.4397651255130768, + "step": 556 + }, + { + "epoch": 0.6363053239255934, + "grad_norm": 90.63822694085678, + "learning_rate": 1.6440816057480812e-07, + "logits/chosen": -1.4811036586761475, + "logits/rejected": -1.554297924041748, + "logps/chosen": -214.304931640625, + "logps/rejected": -267.9236145019531, + "loss": 0.638, + "rewards/accuracies": 0.625, + "rewards/chosen": 0.08694766461849213, + "rewards/margins": 0.4463149309158325, + "rewards/rejected": -0.359367311000824, + "step": 558 + }, + { + "epoch": 0.6385859881690542, + "grad_norm": 109.09749536253133, + "learning_rate": 1.641167532161545e-07, + "logits/chosen": -1.3816839456558228, + "logits/rejected": -1.3854830265045166, + "logps/chosen": -146.0878448486328, + "logps/rejected": -151.40838623046875, + "loss": 0.6199, + "rewards/accuracies": 0.625, + "rewards/chosen": -0.2251240313053131, + "rewards/margins": 0.2835414409637451, + "rewards/rejected": -0.5086655020713806, + "step": 560 + }, + { + "epoch": 0.6408666524125152, + "grad_norm": 85.56418756561474, + "learning_rate": 1.63824418364482e-07, + "logits/chosen": -1.3435323238372803, + "logits/rejected": -1.4011439085006714, + "logps/chosen": -171.78982543945312, + "logps/rejected": -188.12899780273438, + "loss": 0.6342, + "rewards/accuracies": 0.5625, + "rewards/chosen": -0.03885346278548241, + "rewards/margins": 0.2555912137031555, + "rewards/rejected": -0.29444462060928345, + "step": 562 + }, + { + "epoch": 0.643147316655976, + "grad_norm": 91.83203136203487, + "learning_rate": 1.6353116024861583e-07, + "logits/chosen": -1.3896362781524658, + "logits/rejected": -1.3640623092651367, + "logps/chosen": -186.40646362304688, + "logps/rejected": -194.1944580078125, + "loss": 0.6166, + "rewards/accuracies": 0.71875, + "rewards/chosen": -0.14515984058380127, + "rewards/margins": 0.4773363471031189, + "rewards/rejected": -0.6224961876869202, + "step": 564 + }, + { + "epoch": 0.645427980899437, + "grad_norm": 81.53572929971928, + "learning_rate": 1.6323698311073665e-07, + "logits/chosen": -1.2487475872039795, + "logits/rejected": -1.298531413078308, + "logps/chosen": -157.927734375, + "logps/rejected": -172.3856658935547, + "loss": 0.6697, + "rewards/accuracies": 0.65625, + "rewards/chosen": -0.2098311483860016, + "rewards/margins": 0.2927185297012329, + "rewards/rejected": -0.5025496482849121, + "step": 566 + }, + { + "epoch": 0.6477086451428978, + "grad_norm": 114.13313398743495, + "learning_rate": 1.6294189120631955e-07, + "logits/chosen": -1.4274886846542358, + "logits/rejected": -1.5344736576080322, + "logps/chosen": -182.54502868652344, + "logps/rejected": -216.51434326171875, + "loss": 0.6837, + "rewards/accuracies": 0.71875, + "rewards/chosen": -0.11003254354000092, + "rewards/margins": 0.4595148265361786, + "rewards/rejected": -0.5695474147796631, + "step": 568 + }, + { + "epoch": 0.6499893093863588, + "grad_norm": 91.72506895488603, + "learning_rate": 1.6264588880407218e-07, + "logits/chosen": -1.3987553119659424, + "logits/rejected": -1.375367283821106, + "logps/chosen": -150.62136840820312, + "logps/rejected": -147.3133087158203, + "loss": 0.655, + "rewards/accuracies": 0.59375, + "rewards/chosen": 0.012118123471736908, + "rewards/margins": 0.23711130023002625, + "rewards/rejected": -0.22499315440654755, + "step": 570 + }, + { + "epoch": 0.6522699736298196, + "grad_norm": 107.88986883450896, + "learning_rate": 1.6234898018587336e-07, + "logits/chosen": -1.2692500352859497, + "logits/rejected": -1.369061827659607, + "logps/chosen": -139.50912475585938, + "logps/rejected": -160.3131561279297, + "loss": 0.6783, + "rewards/accuracies": 0.46875, + "rewards/chosen": -0.21677826344966888, + "rewards/margins": 0.14994385838508606, + "rewards/rejected": -0.36672213673591614, + "step": 572 + }, + { + "epoch": 0.6545506378732806, + "grad_norm": 78.69743235560887, + "learning_rate": 1.620511696467108e-07, + "logits/chosen": -1.4847999811172485, + "logits/rejected": -1.4455361366271973, + "logps/chosen": -143.21315002441406, + "logps/rejected": -142.0272674560547, + "loss": 0.5931, + "rewards/accuracies": 0.59375, + "rewards/chosen": 0.023255709558725357, + "rewards/margins": 0.20145481824874878, + "rewards/rejected": -0.17819911241531372, + "step": 574 + }, + { + "epoch": 0.6568313021167415, + "grad_norm": 88.95142783400384, + "learning_rate": 1.6175246149461918e-07, + "logits/chosen": -1.2740530967712402, + "logits/rejected": -1.454386830329895, + "logps/chosen": -151.84759521484375, + "logps/rejected": -200.50332641601562, + "loss": 0.6686, + "rewards/accuracies": 0.5, + "rewards/chosen": -0.14912785589694977, + "rewards/margins": 0.19120272994041443, + "rewards/rejected": -0.3403306007385254, + "step": 576 + }, + { + "epoch": 0.6591119663602024, + "grad_norm": 106.08082954445281, + "learning_rate": 1.614528600506178e-07, + "logits/chosen": -1.3865227699279785, + "logits/rejected": -1.375576138496399, + "logps/chosen": -190.82093811035156, + "logps/rejected": -221.74673461914062, + "loss": 0.6078, + "rewards/accuracies": 0.625, + "rewards/chosen": 0.02210266701877117, + "rewards/margins": 0.2864355444908142, + "rewards/rejected": -0.2643328607082367, + "step": 578 + }, + { + "epoch": 0.6613926306036633, + "grad_norm": 87.3395501088531, + "learning_rate": 1.6115236964864796e-07, + "logits/chosen": -1.4399878978729248, + "logits/rejected": -1.4395971298217773, + "logps/chosen": -173.22372436523438, + "logps/rejected": -176.63436889648438, + "loss": 0.6855, + "rewards/accuracies": 0.59375, + "rewards/chosen": -0.08596701920032501, + "rewards/margins": 0.2504124939441681, + "rewards/rejected": -0.3363794982433319, + "step": 580 + }, + { + "epoch": 0.6636732948471242, + "grad_norm": 106.76668867973154, + "learning_rate": 1.6085099463551042e-07, + "logits/chosen": -1.4452544450759888, + "logits/rejected": -1.5641038417816162, + "logps/chosen": -168.6453094482422, + "logps/rejected": -173.80926513671875, + "loss": 0.6725, + "rewards/accuracies": 0.625, + "rewards/chosen": -0.173700749874115, + "rewards/margins": 0.12216023355722427, + "rewards/rejected": -0.29586100578308105, + "step": 582 + }, + { + "epoch": 0.6659539590905851, + "grad_norm": 87.05178703169928, + "learning_rate": 1.6054873937080243e-07, + "logits/chosen": -1.304762601852417, + "logits/rejected": -1.368096113204956, + "logps/chosen": -172.9827423095703, + "logps/rejected": -219.77606201171875, + "loss": 0.6104, + "rewards/accuracies": 0.6875, + "rewards/chosen": -0.15153777599334717, + "rewards/margins": 0.4469776451587677, + "rewards/rejected": -0.5985154509544373, + "step": 584 + }, + { + "epoch": 0.668234623334046, + "grad_norm": 110.06770845486187, + "learning_rate": 1.6024560822685464e-07, + "logits/chosen": -1.2189353704452515, + "logits/rejected": -1.2109813690185547, + "logps/chosen": -124.88150024414062, + "logps/rejected": -134.57229614257812, + "loss": 0.6398, + "rewards/accuracies": 0.59375, + "rewards/chosen": -0.04388846457004547, + "rewards/margins": 0.09341102838516235, + "rewards/rejected": -0.13729947805404663, + "step": 586 + }, + { + "epoch": 0.670515287577507, + "grad_norm": 96.86048631567245, + "learning_rate": 1.5994160558866803e-07, + "logits/chosen": -1.4283928871154785, + "logits/rejected": -1.454673171043396, + "logps/chosen": -141.67201232910156, + "logps/rejected": -165.86553955078125, + "loss": 0.6099, + "rewards/accuracies": 0.75, + "rewards/chosen": -0.005323350429534912, + "rewards/margins": 0.37266457080841064, + "rewards/rejected": -0.37798795104026794, + "step": 588 + }, + { + "epoch": 0.6727959518209679, + "grad_norm": 87.33194066270343, + "learning_rate": 1.5963673585385014e-07, + "logits/chosen": -1.3119065761566162, + "logits/rejected": -1.3611897230148315, + "logps/chosen": -149.3885498046875, + "logps/rejected": -233.38589477539062, + "loss": 0.6465, + "rewards/accuracies": 0.71875, + "rewards/chosen": -0.1451117843389511, + "rewards/margins": 0.5241594314575195, + "rewards/rejected": -0.6692711710929871, + "step": 590 + }, + { + "epoch": 0.6750766160644288, + "grad_norm": 89.80632782962597, + "learning_rate": 1.5933100343255182e-07, + "logits/chosen": -1.478175401687622, + "logits/rejected": -1.5088427066802979, + "logps/chosen": -160.82077026367188, + "logps/rejected": -175.51266479492188, + "loss": 0.6521, + "rewards/accuracies": 0.53125, + "rewards/chosen": -0.026576366275548935, + "rewards/margins": 0.050725750625133514, + "rewards/rejected": -0.07730211317539215, + "step": 592 + }, + { + "epoch": 0.6773572803078897, + "grad_norm": 74.35318626181845, + "learning_rate": 1.5902441274740328e-07, + "logits/chosen": -1.3951764106750488, + "logits/rejected": -1.5363534688949585, + "logps/chosen": -199.1707000732422, + "logps/rejected": -223.73233032226562, + "loss": 0.6397, + "rewards/accuracies": 0.71875, + "rewards/chosen": -0.12182176113128662, + "rewards/margins": 0.23543262481689453, + "rewards/rejected": -0.35725438594818115, + "step": 594 + }, + { + "epoch": 0.6796379445513506, + "grad_norm": 94.79924897563647, + "learning_rate": 1.5871696823344998e-07, + "logits/chosen": -1.296017050743103, + "logits/rejected": -1.3113356828689575, + "logps/chosen": -217.41751098632812, + "logps/rejected": -248.5911865234375, + "loss": 0.6955, + "rewards/accuracies": 0.5, + "rewards/chosen": -0.08776041865348816, + "rewards/margins": -0.076215460896492, + "rewards/rejected": -0.011544965207576752, + "step": 596 + }, + { + "epoch": 0.6819186087948115, + "grad_norm": 92.59481105091255, + "learning_rate": 1.584086743380887e-07, + "logits/chosen": -1.4293127059936523, + "logits/rejected": -1.4433321952819824, + "logps/chosen": -118.54342651367188, + "logps/rejected": -134.85447692871094, + "loss": 0.6368, + "rewards/accuracies": 0.65625, + "rewards/chosen": -0.17175720632076263, + "rewards/margins": 0.14412283897399902, + "rewards/rejected": -0.31588006019592285, + "step": 598 + }, + { + "epoch": 0.6841992730382724, + "grad_norm": 89.66618993893586, + "learning_rate": 1.580995355210031e-07, + "logits/chosen": -1.2930192947387695, + "logits/rejected": -1.368417501449585, + "logps/chosen": -156.90586853027344, + "logps/rejected": -204.53976440429688, + "loss": 0.655, + "rewards/accuracies": 0.59375, + "rewards/chosen": -0.10334792733192444, + "rewards/margins": 0.27799686789512634, + "rewards/rejected": -0.3813447952270508, + "step": 600 + }, + { + "epoch": 0.6841992730382724, + "eval_logits/chosen": -1.4766055345535278, + "eval_logits/rejected": -1.4625221490859985, + "eval_logps/chosen": -128.85867309570312, + "eval_logps/rejected": -132.3487091064453, + "eval_loss": 0.6277530789375305, + "eval_rewards/accuracies": 0.6000000238418579, + "eval_rewards/chosen": 0.10392862558364868, + "eval_rewards/margins": 0.13252468407154083, + "eval_rewards/rejected": -0.028596054762601852, + "eval_runtime": 20.497, + "eval_samples_per_second": 4.879, + "eval_steps_per_second": 1.22, + "step": 600 + }, + { + "epoch": 0.6864799372817333, + "grad_norm": 99.81191522440201, + "learning_rate": 1.577895562540992e-07, + "logits/chosen": -1.5043866634368896, + "logits/rejected": -1.5639445781707764, + "logps/chosen": -176.97195434570312, + "logps/rejected": -200.05955505371094, + "loss": 0.6298, + "rewards/accuracies": 0.5, + "rewards/chosen": -0.08353392779827118, + "rewards/margins": 0.1179167851805687, + "rewards/rejected": -0.20145073533058167, + "step": 602 + }, + { + "epoch": 0.6887606015251942, + "grad_norm": 76.70830865796096, + "learning_rate": 1.574787410214407e-07, + "logits/chosen": -1.409602165222168, + "logits/rejected": -1.4615368843078613, + "logps/chosen": -155.466064453125, + "logps/rejected": -179.92701721191406, + "loss": 0.5929, + "rewards/accuracies": 0.6875, + "rewards/chosen": -0.06697350740432739, + "rewards/margins": 0.43057531118392944, + "rewards/rejected": -0.4975488483905792, + "step": 604 + }, + { + "epoch": 0.6910412657686551, + "grad_norm": 94.3888285857028, + "learning_rate": 1.571670943191841e-07, + "logits/chosen": -1.4215627908706665, + "logits/rejected": -1.4542145729064941, + "logps/chosen": -179.4984130859375, + "logps/rejected": -194.54330444335938, + "loss": 0.6238, + "rewards/accuracies": 0.59375, + "rewards/chosen": -0.06045013293623924, + "rewards/margins": 0.22754909098148346, + "rewards/rejected": -0.2879992127418518, + "step": 606 + }, + { + "epoch": 0.6933219300121161, + "grad_norm": 85.0198174251867, + "learning_rate": 1.5685462065551372e-07, + "logits/chosen": -1.3846269845962524, + "logits/rejected": -1.403570532798767, + "logps/chosen": -119.62281799316406, + "logps/rejected": -122.60442352294922, + "loss": 0.6716, + "rewards/accuracies": 0.5625, + "rewards/chosen": 0.05273966118693352, + "rewards/margins": 0.23132431507110596, + "rewards/rejected": -0.17858465015888214, + "step": 608 + }, + { + "epoch": 0.6956025942555769, + "grad_norm": 106.39373760782738, + "learning_rate": 1.5654132455057648e-07, + "logits/chosen": -1.4965307712554932, + "logits/rejected": -1.5246574878692627, + "logps/chosen": -172.1771240234375, + "logps/rejected": -178.22027587890625, + "loss": 0.6898, + "rewards/accuracies": 0.46875, + "rewards/chosen": -0.24939650297164917, + "rewards/margins": -0.02339550480246544, + "rewards/rejected": -0.22600099444389343, + "step": 610 + }, + { + "epoch": 0.6978832584990379, + "grad_norm": 90.59146954300502, + "learning_rate": 1.562272105364164e-07, + "logits/chosen": -1.2977298498153687, + "logits/rejected": -1.356859803199768, + "logps/chosen": -184.52252197265625, + "logps/rejected": -211.69007873535156, + "loss": 0.6426, + "rewards/accuracies": 0.59375, + "rewards/chosen": -0.14194564521312714, + "rewards/margins": 0.10268077999353409, + "rewards/rejected": -0.24462641775608063, + "step": 612 + }, + { + "epoch": 0.7001639227424987, + "grad_norm": 132.7396200312291, + "learning_rate": 1.5591228315690912e-07, + "logits/chosen": -1.312849760055542, + "logits/rejected": -1.3643529415130615, + "logps/chosen": -183.94964599609375, + "logps/rejected": -215.9215087890625, + "loss": 0.6904, + "rewards/accuracies": 0.65625, + "rewards/chosen": -0.048665039241313934, + "rewards/margins": 0.34853097796440125, + "rewards/rejected": -0.3971959948539734, + "step": 614 + }, + { + "epoch": 0.7024445869859597, + "grad_norm": 85.32152495878141, + "learning_rate": 1.5559654696769627e-07, + "logits/chosen": -1.3156163692474365, + "logits/rejected": -1.3534281253814697, + "logps/chosen": -135.82110595703125, + "logps/rejected": -146.7247314453125, + "loss": 0.6404, + "rewards/accuracies": 0.625, + "rewards/chosen": -0.19835428893566132, + "rewards/margins": 0.1589125692844391, + "rewards/rejected": -0.3572668731212616, + "step": 616 + }, + { + "epoch": 0.7047252512294205, + "grad_norm": 97.72254578285961, + "learning_rate": 1.5528000653611933e-07, + "logits/chosen": -1.3361620903015137, + "logits/rejected": -1.3171348571777344, + "logps/chosen": -133.0448760986328, + "logps/rejected": -157.213623046875, + "loss": 0.6296, + "rewards/accuracies": 0.5625, + "rewards/chosen": -0.2093026041984558, + "rewards/margins": 0.04571903124451637, + "rewards/rejected": -0.2550216317176819, + "step": 618 + }, + { + "epoch": 0.7070059154728815, + "grad_norm": 81.04476758820772, + "learning_rate": 1.5496266644115383e-07, + "logits/chosen": -1.335296869277954, + "logits/rejected": -1.4067720174789429, + "logps/chosen": -166.6826934814453, + "logps/rejected": -201.6751251220703, + "loss": 0.5956, + "rewards/accuracies": 0.6875, + "rewards/chosen": 0.02005193382501602, + "rewards/margins": 0.4108864665031433, + "rewards/rejected": -0.3908345103263855, + "step": 620 + }, + { + "epoch": 0.7092865797163423, + "grad_norm": 117.48862378162204, + "learning_rate": 1.5464453127334292e-07, + "logits/chosen": -1.3234319686889648, + "logits/rejected": -1.3931816816329956, + "logps/chosen": -211.87327575683594, + "logps/rejected": -244.9843292236328, + "loss": 0.6435, + "rewards/accuracies": 0.5625, + "rewards/chosen": -0.23849377036094666, + "rewards/margins": 0.13237711787223816, + "rewards/rejected": -0.3708708584308624, + "step": 622 + }, + { + "epoch": 0.7115672439598033, + "grad_norm": 97.28790849798257, + "learning_rate": 1.5432560563473102e-07, + "logits/chosen": -1.3715243339538574, + "logits/rejected": -1.4115769863128662, + "logps/chosen": -167.96871948242188, + "logps/rejected": -195.65823364257812, + "loss": 0.6435, + "rewards/accuracies": 0.5625, + "rewards/chosen": -0.18149994313716888, + "rewards/margins": 0.21998891234397888, + "rewards/rejected": -0.4014888405799866, + "step": 624 + }, + { + "epoch": 0.7138479082032642, + "grad_norm": 74.77329414265908, + "learning_rate": 1.5400589413879728e-07, + "logits/chosen": -1.3941603899002075, + "logits/rejected": -1.4356334209442139, + "logps/chosen": -146.8162841796875, + "logps/rejected": -166.09603881835938, + "loss": 0.6115, + "rewards/accuracies": 0.71875, + "rewards/chosen": 0.0008982233703136444, + "rewards/margins": 0.2602536678314209, + "rewards/rejected": -0.25935542583465576, + "step": 626 + }, + { + "epoch": 0.7161285724467251, + "grad_norm": 102.40110691278403, + "learning_rate": 1.5368540141038877e-07, + "logits/chosen": -1.33980393409729, + "logits/rejected": -1.3720303773880005, + "logps/chosen": -177.0899658203125, + "logps/rejected": -177.4815216064453, + "loss": 0.6489, + "rewards/accuracies": 0.625, + "rewards/chosen": -0.17625108361244202, + "rewards/margins": 0.22537095844745636, + "rewards/rejected": -0.40162205696105957, + "step": 628 + }, + { + "epoch": 0.718409236690186, + "grad_norm": 77.60752469183791, + "learning_rate": 1.5336413208565372e-07, + "logits/chosen": -1.424604058265686, + "logits/rejected": -1.4951881170272827, + "logps/chosen": -217.2845458984375, + "logps/rejected": -298.6929016113281, + "loss": 0.5522, + "rewards/accuracies": 0.625, + "rewards/chosen": -0.31031134724617004, + "rewards/margins": 0.769706130027771, + "rewards/rejected": -1.0800174474716187, + "step": 630 + }, + { + "epoch": 0.7206899009336469, + "grad_norm": 86.60572317721935, + "learning_rate": 1.5304209081197424e-07, + "logits/chosen": -1.2321975231170654, + "logits/rejected": -1.1972179412841797, + "logps/chosen": -162.22976684570312, + "logps/rejected": -173.07118225097656, + "loss": 0.651, + "rewards/accuracies": 0.6875, + "rewards/chosen": -0.0614384189248085, + "rewards/margins": 0.27045127749443054, + "rewards/rejected": -0.33188968896865845, + "step": 632 + }, + { + "epoch": 0.7229705651771078, + "grad_norm": 105.40040114011227, + "learning_rate": 1.5271928224789933e-07, + "logits/chosen": -1.3975963592529297, + "logits/rejected": -1.4231470823287964, + "logps/chosen": -202.03836059570312, + "logps/rejected": -236.25344848632812, + "loss": 0.6755, + "rewards/accuracies": 0.59375, + "rewards/chosen": -0.32155218720436096, + "rewards/margins": 0.17959336936473846, + "rewards/rejected": -0.5011455416679382, + "step": 634 + }, + { + "epoch": 0.7252512294205687, + "grad_norm": 81.36310946045379, + "learning_rate": 1.5239571106307727e-07, + "logits/chosen": -1.3148503303527832, + "logits/rejected": -1.3296858072280884, + "logps/chosen": -97.38423919677734, + "logps/rejected": -119.96273803710938, + "loss": 0.632, + "rewards/accuracies": 0.6875, + "rewards/chosen": 0.0010671375785022974, + "rewards/margins": 0.2470298409461975, + "rewards/rejected": -0.2459626942873001, + "step": 636 + }, + { + "epoch": 0.7275318936640296, + "grad_norm": 96.83510386866747, + "learning_rate": 1.5207138193818824e-07, + "logits/chosen": -1.4544883966445923, + "logits/rejected": -1.4232516288757324, + "logps/chosen": -173.56967163085938, + "logps/rejected": -202.63575744628906, + "loss": 0.618, + "rewards/accuracies": 0.59375, + "rewards/chosen": -0.11890554428100586, + "rewards/margins": 0.4641531705856323, + "rewards/rejected": -0.5830587148666382, + "step": 638 + }, + { + "epoch": 0.7298125579074906, + "grad_norm": 98.36905363836429, + "learning_rate": 1.5174629956487657e-07, + "logits/chosen": -1.3621355295181274, + "logits/rejected": -1.470801591873169, + "logps/chosen": -152.83348083496094, + "logps/rejected": -180.09095764160156, + "loss": 0.6283, + "rewards/accuracies": 0.5625, + "rewards/chosen": -0.1331956833600998, + "rewards/margins": 0.12644681334495544, + "rewards/rejected": -0.25964248180389404, + "step": 640 + }, + { + "epoch": 0.7320932221509515, + "grad_norm": 99.21748796643946, + "learning_rate": 1.514204686456828e-07, + "logits/chosen": -1.4174845218658447, + "logits/rejected": -1.5134220123291016, + "logps/chosen": -199.45867919921875, + "logps/rejected": -233.53509521484375, + "loss": 0.6424, + "rewards/accuracies": 0.71875, + "rewards/chosen": 0.029536306858062744, + "rewards/margins": 0.40678656101226807, + "rewards/rejected": -0.3772502541542053, + "step": 642 + }, + { + "epoch": 0.7343738863944124, + "grad_norm": 104.85775879115107, + "learning_rate": 1.5109389389397564e-07, + "logits/chosen": -1.309433102607727, + "logits/rejected": -1.3806891441345215, + "logps/chosen": -198.72012329101562, + "logps/rejected": -229.50213623046875, + "loss": 0.636, + "rewards/accuracies": 0.6875, + "rewards/chosen": -0.2979825735092163, + "rewards/margins": 0.2780360281467438, + "rewards/rejected": -0.5760185718536377, + "step": 644 + }, + { + "epoch": 0.7366545506378733, + "grad_norm": 83.70012784480859, + "learning_rate": 1.50766580033884e-07, + "logits/chosen": -1.3813220262527466, + "logits/rejected": -1.4268507957458496, + "logps/chosen": -144.8968048095703, + "logps/rejected": -168.29891967773438, + "loss": 0.6301, + "rewards/accuracies": 0.65625, + "rewards/chosen": -0.2506921887397766, + "rewards/margins": 0.17526662349700928, + "rewards/rejected": -0.4259588122367859, + "step": 646 + }, + { + "epoch": 0.7389352148813342, + "grad_norm": 76.15531374501609, + "learning_rate": 1.5043853180022838e-07, + "logits/chosen": -1.3109767436981201, + "logits/rejected": -1.393604040145874, + "logps/chosen": -123.48320770263672, + "logps/rejected": -146.57994079589844, + "loss": 0.627, + "rewards/accuracies": 0.6875, + "rewards/chosen": -0.030930444598197937, + "rewards/margins": 0.237480029463768, + "rewards/rejected": -0.26841047406196594, + "step": 648 + }, + { + "epoch": 0.7412158791247951, + "grad_norm": 80.96947358910683, + "learning_rate": 1.5010975393845257e-07, + "logits/chosen": -1.249889612197876, + "logits/rejected": -1.2592819929122925, + "logps/chosen": -210.37823486328125, + "logps/rejected": -225.4598388671875, + "loss": 0.6471, + "rewards/accuracies": 0.625, + "rewards/chosen": -0.39817529916763306, + "rewards/margins": 0.08833743631839752, + "rewards/rejected": -0.4865127205848694, + "step": 650 + }, + { + "epoch": 0.743496543368256, + "grad_norm": 87.1921118015952, + "learning_rate": 1.4978025120455482e-07, + "logits/chosen": -1.3601963520050049, + "logits/rejected": -1.376826286315918, + "logps/chosen": -167.55612182617188, + "logps/rejected": -175.45620727539062, + "loss": 0.6323, + "rewards/accuracies": 0.75, + "rewards/chosen": 0.03152412921190262, + "rewards/margins": 0.33548736572265625, + "rewards/rejected": -0.30396321415901184, + "step": 652 + }, + { + "epoch": 0.745777207611717, + "grad_norm": 101.45108780703592, + "learning_rate": 1.4945002836501935e-07, + "logits/chosen": -1.4448864459991455, + "logits/rejected": -1.5169280767440796, + "logps/chosen": -251.17156982421875, + "logps/rejected": -303.9085693359375, + "loss": 0.6505, + "rewards/accuracies": 0.5, + "rewards/chosen": -0.29235783219337463, + "rewards/margins": 0.28852301836013794, + "rewards/rejected": -0.580880880355835, + "step": 654 + }, + { + "epoch": 0.7480578718551778, + "grad_norm": 78.1814307088951, + "learning_rate": 1.4911909019674702e-07, + "logits/chosen": -1.3950117826461792, + "logits/rejected": -1.3936371803283691, + "logps/chosen": -221.95086669921875, + "logps/rejected": -246.12425231933594, + "loss": 0.7095, + "rewards/accuracies": 0.625, + "rewards/chosen": -0.40981170535087585, + "rewards/margins": 0.04004380851984024, + "rewards/rejected": -0.4498555362224579, + "step": 656 + }, + { + "epoch": 0.7503385360986388, + "grad_norm": 86.3290682570179, + "learning_rate": 1.4878744148698654e-07, + "logits/chosen": -1.4538379907608032, + "logits/rejected": -1.4509309530258179, + "logps/chosen": -240.70957946777344, + "logps/rejected": -277.9938049316406, + "loss": 0.6274, + "rewards/accuracies": 0.71875, + "rewards/chosen": -0.10867351293563843, + "rewards/margins": 0.7040268182754517, + "rewards/rejected": -0.8127003908157349, + "step": 658 + }, + { + "epoch": 0.7526192003420996, + "grad_norm": 76.663224249533, + "learning_rate": 1.48455087033265e-07, + "logits/chosen": -1.347740888595581, + "logits/rejected": -1.4315268993377686, + "logps/chosen": -175.90298461914062, + "logps/rejected": -204.00741577148438, + "loss": 0.589, + "rewards/accuracies": 0.8125, + "rewards/chosen": 0.0198584645986557, + "rewards/margins": 0.3997357189655304, + "rewards/rejected": -0.3798772692680359, + "step": 660 + }, + { + "epoch": 0.7548998645855606, + "grad_norm": 89.45613833312, + "learning_rate": 1.4812203164331865e-07, + "logits/chosen": -1.3782517910003662, + "logits/rejected": -1.3579304218292236, + "logps/chosen": -106.40101623535156, + "logps/rejected": -115.47900390625, + "loss": 0.6276, + "rewards/accuracies": 0.6875, + "rewards/chosen": 0.002452205866575241, + "rewards/margins": 0.2672392427921295, + "rewards/rejected": -0.2647870182991028, + "step": 662 + }, + { + "epoch": 0.7571805288290214, + "grad_norm": 90.62698292022132, + "learning_rate": 1.4778828013502315e-07, + "logits/chosen": -1.3181474208831787, + "logits/rejected": -1.2984850406646729, + "logps/chosen": -121.110595703125, + "logps/rejected": -144.5844268798828, + "loss": 0.6345, + "rewards/accuracies": 0.46875, + "rewards/chosen": -0.14049722254276276, + "rewards/margins": 0.16425339877605438, + "rewards/rejected": -0.30475062131881714, + "step": 664 + }, + { + "epoch": 0.7594611930724824, + "grad_norm": 90.46387262398376, + "learning_rate": 1.474538373363241e-07, + "logits/chosen": -1.2038509845733643, + "logits/rejected": -1.1771519184112549, + "logps/chosen": -123.54405212402344, + "logps/rejected": -128.04415893554688, + "loss": 0.6513, + "rewards/accuracies": 0.6875, + "rewards/chosen": -0.0023752544075250626, + "rewards/margins": 0.20540934801101685, + "rewards/rejected": -0.20778462290763855, + "step": 666 + }, + { + "epoch": 0.7617418573159432, + "grad_norm": 98.8814238604518, + "learning_rate": 1.4711870808516705e-07, + "logits/chosen": -1.3800256252288818, + "logits/rejected": -1.4212613105773926, + "logps/chosen": -143.5724639892578, + "logps/rejected": -149.91879272460938, + "loss": 0.6367, + "rewards/accuracies": 0.71875, + "rewards/chosen": -0.25999924540519714, + "rewards/margins": 0.30095675587654114, + "rewards/rejected": -0.5609559416770935, + "step": 668 + }, + { + "epoch": 0.7640225215594042, + "grad_norm": 91.14237505783925, + "learning_rate": 1.4678289722942755e-07, + "logits/chosen": -1.4362088441848755, + "logits/rejected": -1.4652812480926514, + "logps/chosen": -238.3679962158203, + "logps/rejected": -245.81390380859375, + "loss": 0.6248, + "rewards/accuracies": 0.625, + "rewards/chosen": -0.2839330732822418, + "rewards/margins": 0.23081044852733612, + "rewards/rejected": -0.5147435069084167, + "step": 670 + }, + { + "epoch": 0.766303185802865, + "grad_norm": 92.40909256739194, + "learning_rate": 1.4644640962684106e-07, + "logits/chosen": -1.2180765867233276, + "logits/rejected": -1.2128483057022095, + "logps/chosen": -139.85572814941406, + "logps/rejected": -155.68014526367188, + "loss": 0.6594, + "rewards/accuracies": 0.65625, + "rewards/chosen": -0.31525713205337524, + "rewards/margins": 0.2499125599861145, + "rewards/rejected": -0.5651696920394897, + "step": 672 + }, + { + "epoch": 0.768583850046326, + "grad_norm": 85.52183536423232, + "learning_rate": 1.461092501449326e-07, + "logits/chosen": -1.432080864906311, + "logits/rejected": -1.4694151878356934, + "logps/chosen": -192.5015411376953, + "logps/rejected": -206.3730926513672, + "loss": 0.6026, + "rewards/accuracies": 0.6875, + "rewards/chosen": -0.06770830601453781, + "rewards/margins": 0.42051053047180176, + "rewards/rejected": -0.48821884393692017, + "step": 674 + }, + { + "epoch": 0.7708645142897869, + "grad_norm": 90.10321239819011, + "learning_rate": 1.4577142366094641e-07, + "logits/chosen": -1.357291579246521, + "logits/rejected": -1.4703514575958252, + "logps/chosen": -159.84365844726562, + "logps/rejected": -179.19932556152344, + "loss": 0.6417, + "rewards/accuracies": 0.6875, + "rewards/chosen": -0.26335883140563965, + "rewards/margins": 0.19603171944618225, + "rewards/rejected": -0.4593905210494995, + "step": 676 + }, + { + "epoch": 0.7731451785332478, + "grad_norm": 89.71495801422127, + "learning_rate": 1.454329350617754e-07, + "logits/chosen": -1.2440226078033447, + "logits/rejected": -1.2795155048370361, + "logps/chosen": -142.0009765625, + "logps/rejected": -174.45877075195312, + "loss": 0.8425, + "rewards/accuracies": 0.75, + "rewards/chosen": -0.18753905594348907, + "rewards/margins": 0.41140902042388916, + "rewards/rejected": -0.598948061466217, + "step": 678 + }, + { + "epoch": 0.7754258427767087, + "grad_norm": 86.78570750881197, + "learning_rate": 1.4509378924389042e-07, + "logits/chosen": -1.3585668802261353, + "logits/rejected": -1.4263851642608643, + "logps/chosen": -138.0133056640625, + "logps/rejected": -149.3664093017578, + "loss": 0.6293, + "rewards/accuracies": 0.5, + "rewards/chosen": -0.08084426820278168, + "rewards/margins": 0.16473475098609924, + "rewards/rejected": -0.24557898938655853, + "step": 680 + }, + { + "epoch": 0.7777065070201696, + "grad_norm": 107.27075654983396, + "learning_rate": 1.4475399111326942e-07, + "logits/chosen": -1.3274388313293457, + "logits/rejected": -1.4092543125152588, + "logps/chosen": -205.84066772460938, + "logps/rejected": -229.88511657714844, + "loss": 0.6449, + "rewards/accuracies": 0.65625, + "rewards/chosen": -0.28960123658180237, + "rewards/margins": 0.23577138781547546, + "rewards/rejected": -0.5253726243972778, + "step": 682 + }, + { + "epoch": 0.7799871712636305, + "grad_norm": 78.77386252056206, + "learning_rate": 1.4441354558532653e-07, + "logits/chosen": -1.3599066734313965, + "logits/rejected": -1.4156947135925293, + "logps/chosen": -178.33787536621094, + "logps/rejected": -178.77232360839844, + "loss": 0.644, + "rewards/accuracies": 0.6875, + "rewards/chosen": -0.20014306902885437, + "rewards/margins": 0.3175792396068573, + "rewards/rejected": -0.5177222490310669, + "step": 684 + }, + { + "epoch": 0.7822678355070914, + "grad_norm": 96.56956615564326, + "learning_rate": 1.4407245758484092e-07, + "logits/chosen": -1.4600579738616943, + "logits/rejected": -1.4812790155410767, + "logps/chosen": -200.6736602783203, + "logps/rejected": -215.11302185058594, + "loss": 0.6408, + "rewards/accuracies": 0.625, + "rewards/chosen": -0.31072649359703064, + "rewards/margins": 0.2427929937839508, + "rewards/rejected": -0.5535194873809814, + "step": 686 + }, + { + "epoch": 0.7845484997505523, + "grad_norm": 93.7282766427954, + "learning_rate": 1.4373073204588556e-07, + "logits/chosen": -1.392540454864502, + "logits/rejected": -1.5576982498168945, + "logps/chosen": -173.74102783203125, + "logps/rejected": -209.58255004882812, + "loss": 0.6396, + "rewards/accuracies": 0.625, + "rewards/chosen": -0.07411661744117737, + "rewards/margins": 0.2528524100780487, + "rewards/rejected": -0.32696905732154846, + "step": 688 + }, + { + "epoch": 0.7868291639940133, + "grad_norm": 78.03441039978401, + "learning_rate": 1.433883739117558e-07, + "logits/chosen": -1.4191378355026245, + "logits/rejected": -1.376889705657959, + "logps/chosen": -167.5023651123047, + "logps/rejected": -175.2780303955078, + "loss": 0.6232, + "rewards/accuracies": 0.6875, + "rewards/chosen": 0.10266311466693878, + "rewards/margins": 0.33145037293434143, + "rewards/rejected": -0.22878730297088623, + "step": 690 + }, + { + "epoch": 0.7891098282374741, + "grad_norm": 102.68741637989442, + "learning_rate": 1.4304538813489807e-07, + "logits/chosen": -1.365789771080017, + "logits/rejected": -1.4465529918670654, + "logps/chosen": -241.91314697265625, + "logps/rejected": -287.3153381347656, + "loss": 0.626, + "rewards/accuracies": 0.71875, + "rewards/chosen": -0.22250030934810638, + "rewards/margins": 0.3011859655380249, + "rewards/rejected": -0.5236862301826477, + "step": 692 + }, + { + "epoch": 0.7913904924809351, + "grad_norm": 126.01863649654494, + "learning_rate": 1.4270177967683795e-07, + "logits/chosen": -1.1870087385177612, + "logits/rejected": -1.2576615810394287, + "logps/chosen": -193.8482666015625, + "logps/rejected": -214.00135803222656, + "loss": 0.683, + "rewards/accuracies": 0.5625, + "rewards/chosen": -0.6029391884803772, + "rewards/margins": 0.08865418285131454, + "rewards/rejected": -0.69159334897995, + "step": 694 + }, + { + "epoch": 0.793671156724396, + "grad_norm": 105.78445666178517, + "learning_rate": 1.4235755350810852e-07, + "logits/chosen": -1.4039561748504639, + "logits/rejected": -1.483564853668213, + "logps/chosen": -131.8105926513672, + "logps/rejected": -161.2996368408203, + "loss": 0.5968, + "rewards/accuracies": 0.71875, + "rewards/chosen": -0.06916500627994537, + "rewards/margins": 0.4792707562446594, + "rewards/rejected": -0.5484358072280884, + "step": 696 + }, + { + "epoch": 0.7959518209678569, + "grad_norm": 111.82639934813994, + "learning_rate": 1.420127146081786e-07, + "logits/chosen": -1.182243824005127, + "logits/rejected": -1.2272214889526367, + "logps/chosen": -142.89569091796875, + "logps/rejected": -167.72943115234375, + "loss": 0.7077, + "rewards/accuracies": 0.4375, + "rewards/chosen": -0.33521801233291626, + "rewards/margins": 0.031463623046875, + "rewards/rejected": -0.36668163537979126, + "step": 698 + }, + { + "epoch": 0.7982324852113178, + "grad_norm": 93.0448822586256, + "learning_rate": 1.4166726796538043e-07, + "logits/chosen": -1.337456464767456, + "logits/rejected": -1.38818359375, + "logps/chosen": -161.9310760498047, + "logps/rejected": -177.00717163085938, + "loss": 0.5943, + "rewards/accuracies": 0.6875, + "rewards/chosen": -0.11526203155517578, + "rewards/margins": 0.27307793498039246, + "rewards/rejected": -0.38833993673324585, + "step": 700 + }, + { + "epoch": 0.7982324852113178, + "eval_logits/chosen": -1.4622485637664795, + "eval_logits/rejected": -1.4484994411468506, + "eval_logps/chosen": -129.2548065185547, + "eval_logps/rejected": -133.13601684570312, + "eval_loss": 0.6083559989929199, + "eval_rewards/accuracies": 0.6399999856948853, + "eval_rewards/chosen": 0.06431641429662704, + "eval_rewards/margins": 0.17164374887943268, + "eval_rewards/rejected": -0.10732734948396683, + "eval_runtime": 19.822, + "eval_samples_per_second": 5.045, + "eval_steps_per_second": 1.261, + "step": 700 + }, + { + "epoch": 0.8005131494547787, + "grad_norm": 81.54591894736863, + "learning_rate": 1.413212185768378e-07, + "logits/chosen": -1.3647618293762207, + "logits/rejected": -1.4832388162612915, + "logps/chosen": -163.0651092529297, + "logps/rejected": -199.4841766357422, + "loss": 0.6247, + "rewards/accuracies": 0.8125, + "rewards/chosen": 0.14452999830245972, + "rewards/margins": 0.4548623561859131, + "rewards/rejected": -0.31033238768577576, + "step": 702 + }, + { + "epoch": 0.8027938136982397, + "grad_norm": 88.3455225094938, + "learning_rate": 1.409745714483936e-07, + "logits/chosen": -1.3221628665924072, + "logits/rejected": -1.379399299621582, + "logps/chosen": -124.75774383544922, + "logps/rejected": -142.49942016601562, + "loss": 0.6477, + "rewards/accuracies": 0.65625, + "rewards/chosen": -0.004875157028436661, + "rewards/margins": 0.23325148224830627, + "rewards/rejected": -0.23812663555145264, + "step": 704 + }, + { + "epoch": 0.8050744779417005, + "grad_norm": 104.14495646431855, + "learning_rate": 1.406273315945374e-07, + "logits/chosen": -1.4262161254882812, + "logits/rejected": -1.460605502128601, + "logps/chosen": -188.75218200683594, + "logps/rejected": -196.2736358642578, + "loss": 0.6307, + "rewards/accuracies": 0.625, + "rewards/chosen": -0.09237577021121979, + "rewards/margins": 0.26734021306037903, + "rewards/rejected": -0.35971593856811523, + "step": 706 + }, + { + "epoch": 0.8073551421851615, + "grad_norm": 88.27883050256283, + "learning_rate": 1.4027950403833294e-07, + "logits/chosen": -1.3447182178497314, + "logits/rejected": -1.3849260807037354, + "logps/chosen": -194.97628784179688, + "logps/rejected": -231.31488037109375, + "loss": 0.608, + "rewards/accuracies": 0.65625, + "rewards/chosen": -0.018238365650177002, + "rewards/margins": 0.4262167513370514, + "rewards/rejected": -0.444455087184906, + "step": 708 + }, + { + "epoch": 0.8096358064286223, + "grad_norm": 83.81720367834012, + "learning_rate": 1.3993109381134552e-07, + "logits/chosen": -1.458938717842102, + "logits/rejected": -1.4975433349609375, + "logps/chosen": -217.40234375, + "logps/rejected": -230.56849670410156, + "loss": 0.5997, + "rewards/accuracies": 0.6875, + "rewards/chosen": -0.06927414983510971, + "rewards/margins": 0.34377092123031616, + "rewards/rejected": -0.41304510831832886, + "step": 710 + }, + { + "epoch": 0.8119164706720833, + "grad_norm": 83.92843092918012, + "learning_rate": 1.3958210595356923e-07, + "logits/chosen": -1.295701265335083, + "logits/rejected": -1.3689717054367065, + "logps/chosen": -184.71035766601562, + "logps/rejected": -231.5858154296875, + "loss": 0.6496, + "rewards/accuracies": 0.75, + "rewards/chosen": -0.07217526435852051, + "rewards/margins": 0.5205077528953552, + "rewards/rejected": -0.5926830768585205, + "step": 712 + }, + { + "epoch": 0.8141971349155441, + "grad_norm": 64.10905806570788, + "learning_rate": 1.3923254551335385e-07, + "logits/chosen": -1.3552354574203491, + "logits/rejected": -1.5121649503707886, + "logps/chosen": -162.65443420410156, + "logps/rejected": -202.2783203125, + "loss": 0.6038, + "rewards/accuracies": 0.65625, + "rewards/chosen": -0.21346963942050934, + "rewards/margins": 0.2602553367614746, + "rewards/rejected": -0.47372499108314514, + "step": 714 + }, + { + "epoch": 0.8164777991590051, + "grad_norm": 96.65072480066111, + "learning_rate": 1.388824175473321e-07, + "logits/chosen": -1.4826998710632324, + "logits/rejected": -1.525536060333252, + "logps/chosen": -157.91253662109375, + "logps/rejected": -189.42970275878906, + "loss": 0.6418, + "rewards/accuracies": 0.5625, + "rewards/chosen": -0.15659984946250916, + "rewards/margins": 0.5801843404769897, + "rewards/rejected": -0.7367842197418213, + "step": 716 + }, + { + "epoch": 0.8187584634024659, + "grad_norm": 133.38988345223416, + "learning_rate": 1.3853172712034618e-07, + "logits/chosen": -1.453382968902588, + "logits/rejected": -1.4554516077041626, + "logps/chosen": -123.51729583740234, + "logps/rejected": -133.98687744140625, + "loss": 0.6316, + "rewards/accuracies": 0.5625, + "rewards/chosen": -0.17259877920150757, + "rewards/margins": 0.09528855979442596, + "rewards/rejected": -0.26788732409477234, + "step": 718 + }, + { + "epoch": 0.8210391276459269, + "grad_norm": 103.6833810750344, + "learning_rate": 1.3818047930537489e-07, + "logits/chosen": -1.4238344430923462, + "logits/rejected": -1.4430863857269287, + "logps/chosen": -185.29090881347656, + "logps/rejected": -190.6168212890625, + "loss": 0.6817, + "rewards/accuracies": 0.5, + "rewards/chosen": -0.1492576152086258, + "rewards/margins": 0.1548936814069748, + "rewards/rejected": -0.304151326417923, + "step": 720 + }, + { + "epoch": 0.8233197918893878, + "grad_norm": 104.78480311225195, + "learning_rate": 1.3782867918345986e-07, + "logits/chosen": -1.4426988363265991, + "logits/rejected": -1.474928855895996, + "logps/chosen": -182.13693237304688, + "logps/rejected": -207.8540802001953, + "loss": 0.688, + "rewards/accuracies": 0.5625, + "rewards/chosen": -0.28544071316719055, + "rewards/margins": 0.010512780398130417, + "rewards/rejected": -0.2959534525871277, + "step": 722 + }, + { + "epoch": 0.8256004561328487, + "grad_norm": 83.99721091366573, + "learning_rate": 1.374763318436323e-07, + "logits/chosen": -1.3846015930175781, + "logits/rejected": -1.374580979347229, + "logps/chosen": -188.639892578125, + "logps/rejected": -203.4746856689453, + "loss": 0.6455, + "rewards/accuracies": 0.59375, + "rewards/chosen": -0.09084916859865189, + "rewards/margins": 0.13614203035831451, + "rewards/rejected": -0.2269912213087082, + "step": 724 + }, + { + "epoch": 0.8278811203763096, + "grad_norm": 100.80568473474456, + "learning_rate": 1.371234423828393e-07, + "logits/chosen": -1.3973044157028198, + "logits/rejected": -1.4167506694793701, + "logps/chosen": -221.90057373046875, + "logps/rejected": -223.55625915527344, + "loss": 0.6358, + "rewards/accuracies": 0.5, + "rewards/chosen": -0.5376750826835632, + "rewards/margins": 0.1100195124745369, + "rewards/rejected": -0.6476945281028748, + "step": 726 + }, + { + "epoch": 0.8301617846197705, + "grad_norm": 95.2695276072168, + "learning_rate": 1.367700159058701e-07, + "logits/chosen": -1.304479718208313, + "logits/rejected": -1.3024879693984985, + "logps/chosen": -132.629150390625, + "logps/rejected": -157.943115234375, + "loss": 0.6385, + "rewards/accuracies": 0.625, + "rewards/chosen": -0.17815637588500977, + "rewards/margins": 0.16010738909244537, + "rewards/rejected": -0.33826377987861633, + "step": 728 + }, + { + "epoch": 0.8324424488632314, + "grad_norm": 84.57675708069675, + "learning_rate": 1.3641605752528223e-07, + "logits/chosen": -1.345677137374878, + "logits/rejected": -1.369272232055664, + "logps/chosen": -174.67979431152344, + "logps/rejected": -177.95265197753906, + "loss": 0.6631, + "rewards/accuracies": 0.375, + "rewards/chosen": -0.33565065264701843, + "rewards/margins": -0.10611479729413986, + "rewards/rejected": -0.22953587770462036, + "step": 730 + }, + { + "epoch": 0.8347231131066923, + "grad_norm": 95.98437315557469, + "learning_rate": 1.3606157236132753e-07, + "logits/chosen": -1.3388769626617432, + "logits/rejected": -1.4168953895568848, + "logps/chosen": -128.86917114257812, + "logps/rejected": -155.26756286621094, + "loss": 0.5995, + "rewards/accuracies": 0.65625, + "rewards/chosen": -0.12645135819911957, + "rewards/margins": 0.4350703954696655, + "rewards/rejected": -0.5615217685699463, + "step": 732 + }, + { + "epoch": 0.8370037773501532, + "grad_norm": 98.72975210848782, + "learning_rate": 1.3570656554187823e-07, + "logits/chosen": -1.293025255203247, + "logits/rejected": -1.3117451667785645, + "logps/chosen": -198.599609375, + "logps/rejected": -195.0568084716797, + "loss": 0.6664, + "rewards/accuracies": 0.53125, + "rewards/chosen": -0.3894595205783844, + "rewards/margins": 0.01842033676803112, + "rewards/rejected": -0.40787985920906067, + "step": 734 + }, + { + "epoch": 0.8392844415936142, + "grad_norm": 83.44989797339352, + "learning_rate": 1.353510422023526e-07, + "logits/chosen": -1.3290454149246216, + "logits/rejected": -1.3762670755386353, + "logps/chosen": -108.33480834960938, + "logps/rejected": -132.43887329101562, + "loss": 0.6699, + "rewards/accuracies": 0.625, + "rewards/chosen": -0.07590903341770172, + "rewards/margins": 0.31655895709991455, + "rewards/rejected": -0.3924679458141327, + "step": 736 + }, + { + "epoch": 0.841565105837075, + "grad_norm": 94.8310467886239, + "learning_rate": 1.3499500748564075e-07, + "logits/chosen": -1.4289618730545044, + "logits/rejected": -1.4254275560379028, + "logps/chosen": -111.20790100097656, + "logps/rejected": -120.1327896118164, + "loss": 0.6651, + "rewards/accuracies": 0.5, + "rewards/chosen": -0.06255226582288742, + "rewards/margins": 0.16382303833961487, + "rewards/rejected": -0.2263752967119217, + "step": 738 + }, + { + "epoch": 0.843845770080536, + "grad_norm": 106.55945054665669, + "learning_rate": 1.346384665420302e-07, + "logits/chosen": -1.2947108745574951, + "logits/rejected": -1.3571364879608154, + "logps/chosen": -157.76895141601562, + "logps/rejected": -187.6517333984375, + "loss": 0.6321, + "rewards/accuracies": 0.75, + "rewards/chosen": -0.14570750296115875, + "rewards/margins": 0.39257413148880005, + "rewards/rejected": -0.5382816791534424, + "step": 740 + }, + { + "epoch": 0.8461264343239968, + "grad_norm": 112.76083171183551, + "learning_rate": 1.3428142452913133e-07, + "logits/chosen": -1.4128669500350952, + "logits/rejected": -1.4613770246505737, + "logps/chosen": -202.3125457763672, + "logps/rejected": -242.2345428466797, + "loss": 0.6542, + "rewards/accuracies": 0.6875, + "rewards/chosen": -0.27778810262680054, + "rewards/margins": 0.3144941031932831, + "rewards/rejected": -0.592282235622406, + "step": 742 + }, + { + "epoch": 0.8484070985674578, + "grad_norm": 100.11503458436644, + "learning_rate": 1.3392388661180302e-07, + "logits/chosen": -1.3573894500732422, + "logits/rejected": -1.344806432723999, + "logps/chosen": -186.35108947753906, + "logps/rejected": -188.49490356445312, + "loss": 0.6701, + "rewards/accuracies": 0.5625, + "rewards/chosen": -0.16072505712509155, + "rewards/margins": 0.19017082452774048, + "rewards/rejected": -0.3508959114551544, + "step": 744 + }, + { + "epoch": 0.8506877628109187, + "grad_norm": 91.60219394383267, + "learning_rate": 1.3356585796207755e-07, + "logits/chosen": -1.3965177536010742, + "logits/rejected": -1.4244239330291748, + "logps/chosen": -201.50318908691406, + "logps/rejected": -217.1870574951172, + "loss": 0.6803, + "rewards/accuracies": 0.6875, + "rewards/chosen": 0.007203985005617142, + "rewards/margins": 0.4475135803222656, + "rewards/rejected": -0.44030967354774475, + "step": 746 + }, + { + "epoch": 0.8529684270543796, + "grad_norm": 93.27202711374441, + "learning_rate": 1.3320734375908606e-07, + "logits/chosen": -1.4266613721847534, + "logits/rejected": -1.4168901443481445, + "logps/chosen": -208.13624572753906, + "logps/rejected": -238.91502380371094, + "loss": 0.6581, + "rewards/accuracies": 0.65625, + "rewards/chosen": -0.2992842197418213, + "rewards/margins": 0.2575336992740631, + "rewards/rejected": -0.556817889213562, + "step": 748 + }, + { + "epoch": 0.8552490912978405, + "grad_norm": 99.8766408873739, + "learning_rate": 1.328483491889836e-07, + "logits/chosen": -1.427262306213379, + "logits/rejected": -1.4984745979309082, + "logps/chosen": -215.9974365234375, + "logps/rejected": -236.14759826660156, + "loss": 0.6592, + "rewards/accuracies": 0.59375, + "rewards/chosen": -0.048569947481155396, + "rewards/margins": 0.15008097887039185, + "rewards/rejected": -0.19865092635154724, + "step": 750 + }, + { + "epoch": 0.8575297555413014, + "grad_norm": 82.7550556411054, + "learning_rate": 1.3248887944487393e-07, + "logits/chosen": -1.410501480102539, + "logits/rejected": -1.452683687210083, + "logps/chosen": -122.24147033691406, + "logps/rejected": -137.63116455078125, + "loss": 0.6134, + "rewards/accuracies": 0.6875, + "rewards/chosen": 0.11725883185863495, + "rewards/margins": 0.1699836403131485, + "rewards/rejected": -0.05272480472922325, + "step": 752 + }, + { + "epoch": 0.8598104197847624, + "grad_norm": 88.9952584850963, + "learning_rate": 1.321289397267347e-07, + "logits/chosen": -1.4281740188598633, + "logits/rejected": -1.445003628730774, + "logps/chosen": -177.8570556640625, + "logps/rejected": -181.70663452148438, + "loss": 0.6919, + "rewards/accuracies": 0.5625, + "rewards/chosen": -0.2182515561580658, + "rewards/margins": 0.0669536367058754, + "rewards/rejected": -0.28520524501800537, + "step": 754 + }, + { + "epoch": 0.8620910840282232, + "grad_norm": 100.22335380917765, + "learning_rate": 1.3176853524134198e-07, + "logits/chosen": -1.391309142112732, + "logits/rejected": -1.4018808603286743, + "logps/chosen": -164.9562530517578, + "logps/rejected": -178.225341796875, + "loss": 0.6706, + "rewards/accuracies": 0.59375, + "rewards/chosen": -0.04197298735380173, + "rewards/margins": 0.15188831090927124, + "rewards/rejected": -0.19386129081249237, + "step": 756 + }, + { + "epoch": 0.8643717482716842, + "grad_norm": 100.1727346446802, + "learning_rate": 1.314076712021949e-07, + "logits/chosen": -1.35850989818573, + "logits/rejected": -1.420215129852295, + "logps/chosen": -157.33843994140625, + "logps/rejected": -192.61769104003906, + "loss": 0.6228, + "rewards/accuracies": 0.625, + "rewards/chosen": -0.10978934168815613, + "rewards/margins": 0.24940092861652374, + "rewards/rejected": -0.3591902256011963, + "step": 758 + }, + { + "epoch": 0.866652412515145, + "grad_norm": 89.59203114100742, + "learning_rate": 1.3104635282944052e-07, + "logits/chosen": -1.4790459871292114, + "logits/rejected": -1.4666205644607544, + "logps/chosen": -161.86483764648438, + "logps/rejected": -211.35423278808594, + "loss": 0.6404, + "rewards/accuracies": 0.71875, + "rewards/chosen": 0.004301717504858971, + "rewards/margins": 0.35839876532554626, + "rewards/rejected": -0.35409703850746155, + "step": 760 + }, + { + "epoch": 0.868933076758606, + "grad_norm": 94.6581914253334, + "learning_rate": 1.3068458534979812e-07, + "logits/chosen": -1.3811688423156738, + "logits/rejected": -1.4190725088119507, + "logps/chosen": -80.00721740722656, + "logps/rejected": -101.73347473144531, + "loss": 0.6688, + "rewards/accuracies": 0.625, + "rewards/chosen": -0.02035554125905037, + "rewards/margins": 0.07091270387172699, + "rewards/rejected": -0.09126824140548706, + "step": 762 + }, + { + "epoch": 0.8712137410020668, + "grad_norm": 84.05822092329534, + "learning_rate": 1.3032237399648357e-07, + "logits/chosen": -1.3661377429962158, + "logits/rejected": -1.3815851211547852, + "logps/chosen": -106.82656860351562, + "logps/rejected": -128.75656127929688, + "loss": 0.6309, + "rewards/accuracies": 0.5625, + "rewards/chosen": -0.03877873718738556, + "rewards/margins": 0.08257640898227692, + "rewards/rejected": -0.12135513871908188, + "step": 764 + }, + { + "epoch": 0.8734944052455278, + "grad_norm": 99.58060667623182, + "learning_rate": 1.2995972400913367e-07, + "logits/chosen": -1.4144562482833862, + "logits/rejected": -1.4412004947662354, + "logps/chosen": -136.15481567382812, + "logps/rejected": -141.8118438720703, + "loss": 0.6589, + "rewards/accuracies": 0.65625, + "rewards/chosen": -0.0782175287604332, + "rewards/margins": 0.17598725855350494, + "rewards/rejected": -0.25420477986335754, + "step": 766 + }, + { + "epoch": 0.8757750694889886, + "grad_norm": 80.17753876582776, + "learning_rate": 1.2959664063373042e-07, + "logits/chosen": -1.3012042045593262, + "logits/rejected": -1.2482383251190186, + "logps/chosen": -100.65060424804688, + "logps/rejected": -112.65821838378906, + "loss": 0.5368, + "rewards/accuracies": 0.78125, + "rewards/chosen": 0.04776890203356743, + "rewards/margins": 0.4462125897407532, + "rewards/rejected": -0.39844369888305664, + "step": 768 + }, + { + "epoch": 0.8780557337324496, + "grad_norm": 88.16104383702894, + "learning_rate": 1.2923312912252506e-07, + "logits/chosen": -1.4704458713531494, + "logits/rejected": -1.5337769985198975, + "logps/chosen": -118.58712768554688, + "logps/rejected": -128.90980529785156, + "loss": 0.6971, + "rewards/accuracies": 0.53125, + "rewards/chosen": 0.005523152183741331, + "rewards/margins": 0.031648553907871246, + "rewards/rejected": -0.02612539939582348, + "step": 770 + }, + { + "epoch": 0.8803363979759105, + "grad_norm": 88.57684727206627, + "learning_rate": 1.288691947339621e-07, + "logits/chosen": -1.4641995429992676, + "logits/rejected": -1.4404255151748657, + "logps/chosen": -188.07818603515625, + "logps/rejected": -189.4072265625, + "loss": 0.6435, + "rewards/accuracies": 0.65625, + "rewards/chosen": -0.11661610007286072, + "rewards/margins": 0.22115452587604523, + "rewards/rejected": -0.33777061104774475, + "step": 772 + }, + { + "epoch": 0.8826170622193714, + "grad_norm": 88.03404600894119, + "learning_rate": 1.2850484273260325e-07, + "logits/chosen": -1.2959342002868652, + "logits/rejected": -1.3783100843429565, + "logps/chosen": -205.9984130859375, + "logps/rejected": -227.6925506591797, + "loss": 0.6664, + "rewards/accuracies": 0.59375, + "rewards/chosen": -0.16467887163162231, + "rewards/margins": 0.33329781889915466, + "rewards/rejected": -0.497976690530777, + "step": 774 + }, + { + "epoch": 0.8848977264628323, + "grad_norm": 85.2906054415725, + "learning_rate": 1.2814007838905129e-07, + "logits/chosen": -1.3777484893798828, + "logits/rejected": -1.4851555824279785, + "logps/chosen": -169.7268524169922, + "logps/rejected": -183.84872436523438, + "loss": 0.5976, + "rewards/accuracies": 0.65625, + "rewards/chosen": -0.050482071936130524, + "rewards/margins": 0.21894827485084534, + "rewards/rejected": -0.26943036913871765, + "step": 776 + }, + { + "epoch": 0.8871783907062932, + "grad_norm": 113.67510179823725, + "learning_rate": 1.2777490697987375e-07, + "logits/chosen": -1.361803412437439, + "logits/rejected": -1.4689048528671265, + "logps/chosen": -188.6908416748047, + "logps/rejected": -239.35243225097656, + "loss": 0.6593, + "rewards/accuracies": 0.71875, + "rewards/chosen": 0.024178601801395416, + "rewards/margins": 0.5260721445083618, + "rewards/rejected": -0.5018935203552246, + "step": 778 + }, + { + "epoch": 0.8894590549497541, + "grad_norm": 77.93754881822308, + "learning_rate": 1.2740933378752683e-07, + "logits/chosen": -1.4177353382110596, + "logits/rejected": -1.4752013683319092, + "logps/chosen": -196.52670288085938, + "logps/rejected": -216.70066833496094, + "loss": 0.5885, + "rewards/accuracies": 0.75, + "rewards/chosen": -0.18388278782367706, + "rewards/margins": 0.34562817215919495, + "rewards/rejected": -0.5295109748840332, + "step": 780 + }, + { + "epoch": 0.891739719193215, + "grad_norm": 85.24959090845118, + "learning_rate": 1.2704336410027862e-07, + "logits/chosen": -1.4634852409362793, + "logits/rejected": -1.5205847024917603, + "logps/chosen": -250.54759216308594, + "logps/rejected": -285.6510009765625, + "loss": 0.6189, + "rewards/accuracies": 0.53125, + "rewards/chosen": -0.32170242071151733, + "rewards/margins": 0.21656639873981476, + "rewards/rejected": -0.5382688045501709, + "step": 782 + }, + { + "epoch": 0.8940203834366759, + "grad_norm": 94.62957856064564, + "learning_rate": 1.2667700321213279e-07, + "logits/chosen": -1.216140627861023, + "logits/rejected": -1.3199231624603271, + "logps/chosen": -163.63653564453125, + "logps/rejected": -182.559814453125, + "loss": 0.6221, + "rewards/accuracies": 0.53125, + "rewards/chosen": -0.21417541801929474, + "rewards/margins": 0.07622166723012924, + "rewards/rejected": -0.29039713740348816, + "step": 784 + }, + { + "epoch": 0.8963010476801369, + "grad_norm": 84.13982267155843, + "learning_rate": 1.263102564227521e-07, + "logits/chosen": -1.3924446105957031, + "logits/rejected": -1.3973028659820557, + "logps/chosen": -164.72738647460938, + "logps/rejected": -185.91358947753906, + "loss": 0.6378, + "rewards/accuracies": 0.6875, + "rewards/chosen": -0.017229687422513962, + "rewards/margins": 0.29563748836517334, + "rewards/rejected": -0.3128671646118164, + "step": 786 + }, + { + "epoch": 0.8985817119235977, + "grad_norm": 92.91613247214089, + "learning_rate": 1.2594312903738162e-07, + "logits/chosen": -1.4232394695281982, + "logits/rejected": -1.5436615943908691, + "logps/chosen": -137.60520935058594, + "logps/rejected": -159.5936737060547, + "loss": 0.6099, + "rewards/accuracies": 0.6875, + "rewards/chosen": -0.14493076503276825, + "rewards/margins": 0.23380112648010254, + "rewards/rejected": -0.3787318468093872, + "step": 788 + }, + { + "epoch": 0.9008623761670587, + "grad_norm": 109.24475281169839, + "learning_rate": 1.2557562636677192e-07, + "logits/chosen": -1.3665200471878052, + "logits/rejected": -1.4350488185882568, + "logps/chosen": -191.39637756347656, + "logps/rejected": -255.76321411132812, + "loss": 0.6655, + "rewards/accuracies": 0.6875, + "rewards/chosen": -0.16933086514472961, + "rewards/margins": 0.5223888158798218, + "rewards/rejected": -0.691719651222229, + "step": 790 + }, + { + "epoch": 0.9031430404105195, + "grad_norm": 66.209268910187, + "learning_rate": 1.252077537271024e-07, + "logits/chosen": -1.318701982498169, + "logits/rejected": -1.3234184980392456, + "logps/chosen": -180.39767456054688, + "logps/rejected": -182.42808532714844, + "loss": 0.5986, + "rewards/accuracies": 0.71875, + "rewards/chosen": -0.012874465435743332, + "rewards/margins": 0.4246137738227844, + "rewards/rejected": -0.43748825788497925, + "step": 792 + }, + { + "epoch": 0.9054237046539805, + "grad_norm": 110.64800895876068, + "learning_rate": 1.2483951643990425e-07, + "logits/chosen": -1.427907943725586, + "logits/rejected": -1.4352999925613403, + "logps/chosen": -234.02430725097656, + "logps/rejected": -252.70291137695312, + "loss": 0.6882, + "rewards/accuracies": 0.59375, + "rewards/chosen": -0.23474670946598053, + "rewards/margins": 0.3372371196746826, + "rewards/rejected": -0.571983814239502, + "step": 794 + }, + { + "epoch": 0.9077043688974413, + "grad_norm": 71.22024881783034, + "learning_rate": 1.2447091983198366e-07, + "logits/chosen": -1.4595139026641846, + "logits/rejected": -1.494476556777954, + "logps/chosen": -177.576904296875, + "logps/rejected": -200.62152099609375, + "loss": 0.622, + "rewards/accuracies": 0.6875, + "rewards/chosen": -0.10567592829465866, + "rewards/margins": 0.4641418755054474, + "rewards/rejected": -0.5698177814483643, + "step": 796 + }, + { + "epoch": 0.9099850331409023, + "grad_norm": 77.66730114746076, + "learning_rate": 1.2410196923534453e-07, + "logits/chosen": -1.3751342296600342, + "logits/rejected": -1.4557826519012451, + "logps/chosen": -156.71981811523438, + "logps/rejected": -175.78079223632812, + "loss": 0.6241, + "rewards/accuracies": 0.5625, + "rewards/chosen": -0.17370270192623138, + "rewards/margins": 0.5142669081687927, + "rewards/rejected": -0.6879696249961853, + "step": 798 + }, + { + "epoch": 0.9122656973843632, + "grad_norm": 79.01703720689792, + "learning_rate": 1.237326699871115e-07, + "logits/chosen": -1.3648015260696411, + "logits/rejected": -1.4242901802062988, + "logps/chosen": -139.64234924316406, + "logps/rejected": -162.11544799804688, + "loss": 0.6048, + "rewards/accuracies": 0.65625, + "rewards/chosen": -0.2178276777267456, + "rewards/margins": 0.2655281722545624, + "rewards/rejected": -0.48335587978363037, + "step": 800 + }, + { + "epoch": 0.9122656973843632, + "eval_logits/chosen": -1.4735329151153564, + "eval_logits/rejected": -1.460737705230713, + "eval_logps/chosen": -128.9962158203125, + "eval_logps/rejected": -133.23785400390625, + "eval_loss": 0.6001502871513367, + "eval_rewards/accuracies": 0.6800000071525574, + "eval_rewards/chosen": 0.09017454832792282, + "eval_rewards/margins": 0.20768418908119202, + "eval_rewards/rejected": -0.11750967055559158, + "eval_runtime": 19.4284, + "eval_samples_per_second": 5.147, + "eval_steps_per_second": 1.287, + "step": 800 + }, + { + "epoch": 0.9145463616278241, + "grad_norm": 114.49291402716334, + "learning_rate": 1.2336302742945265e-07, + "logits/chosen": -1.1926456689834595, + "logits/rejected": -1.1694424152374268, + "logps/chosen": -151.49020385742188, + "logps/rejected": -221.17051696777344, + "loss": 0.6541, + "rewards/accuracies": 0.4375, + "rewards/chosen": -0.3034975230693817, + "rewards/margins": 0.10761985182762146, + "rewards/rejected": -0.41111740469932556, + "step": 802 + }, + { + "epoch": 0.9168270258712851, + "grad_norm": 87.25759889144325, + "learning_rate": 1.2299304690950234e-07, + "logits/chosen": -1.314978837966919, + "logits/rejected": -1.3282618522644043, + "logps/chosen": -130.7468719482422, + "logps/rejected": -169.4188690185547, + "loss": 0.6349, + "rewards/accuracies": 0.5625, + "rewards/chosen": -0.07509980350732803, + "rewards/margins": 0.23689687252044678, + "rewards/rejected": -0.3119966685771942, + "step": 804 + }, + { + "epoch": 0.9191076901147459, + "grad_norm": 90.2755897487053, + "learning_rate": 1.2262273377928374e-07, + "logits/chosen": -1.4546884298324585, + "logits/rejected": -1.5022200345993042, + "logps/chosen": -155.28553771972656, + "logps/rejected": -186.28756713867188, + "loss": 0.6376, + "rewards/accuracies": 0.5, + "rewards/chosen": -0.20369568467140198, + "rewards/margins": 0.15442761778831482, + "rewards/rejected": -0.3581232726573944, + "step": 806 + }, + { + "epoch": 0.9213883543582069, + "grad_norm": 87.61957852766155, + "learning_rate": 1.2225209339563143e-07, + "logits/chosen": -1.1901319026947021, + "logits/rejected": -1.2732369899749756, + "logps/chosen": -126.2178726196289, + "logps/rejected": -131.4507598876953, + "loss": 0.6119, + "rewards/accuracies": 0.71875, + "rewards/chosen": -0.06662334501743317, + "rewards/margins": 0.1810164749622345, + "rewards/rejected": -0.24763983488082886, + "step": 808 + }, + { + "epoch": 0.9236690186016677, + "grad_norm": 123.86797719697702, + "learning_rate": 1.2188113112011404e-07, + "logits/chosen": -1.260641098022461, + "logits/rejected": -1.3186612129211426, + "logps/chosen": -247.8528594970703, + "logps/rejected": -279.6898498535156, + "loss": 0.6643, + "rewards/accuracies": 0.6875, + "rewards/chosen": -0.3332889676094055, + "rewards/margins": 0.2615528106689453, + "rewards/rejected": -0.594841718673706, + "step": 810 + }, + { + "epoch": 0.9259496828451287, + "grad_norm": 92.05739920896082, + "learning_rate": 1.2150985231895645e-07, + "logits/chosen": -1.5352187156677246, + "logits/rejected": -1.5089805126190186, + "logps/chosen": -168.19351196289062, + "logps/rejected": -155.74728393554688, + "loss": 0.6464, + "rewards/accuracies": 0.65625, + "rewards/chosen": -0.05093672126531601, + "rewards/margins": 0.23030051589012146, + "rewards/rejected": -0.2812372148036957, + "step": 812 + }, + { + "epoch": 0.9282303470885895, + "grad_norm": 93.01772334051867, + "learning_rate": 1.2113826236296244e-07, + "logits/chosen": -1.4838998317718506, + "logits/rejected": -1.4913954734802246, + "logps/chosen": -145.08023071289062, + "logps/rejected": -237.1237030029297, + "loss": 0.63, + "rewards/accuracies": 0.5625, + "rewards/chosen": -0.10176398605108261, + "rewards/margins": 0.37156644463539124, + "rewards/rejected": -0.47333040833473206, + "step": 814 + }, + { + "epoch": 0.9305110113320505, + "grad_norm": 104.77223934596387, + "learning_rate": 1.207663666274367e-07, + "logits/chosen": -1.3763738870620728, + "logits/rejected": -1.3552130460739136, + "logps/chosen": -177.25376892089844, + "logps/rejected": -221.85498046875, + "loss": 0.6857, + "rewards/accuracies": 0.5, + "rewards/chosen": -0.18679024279117584, + "rewards/margins": 0.05987125262618065, + "rewards/rejected": -0.2466614842414856, + "step": 816 + }, + { + "epoch": 0.9327916755755113, + "grad_norm": 103.0413802693114, + "learning_rate": 1.2039417049210742e-07, + "logits/chosen": -1.3318116664886475, + "logits/rejected": -1.3942524194717407, + "logps/chosen": -142.11920166015625, + "logps/rejected": -182.98092651367188, + "loss": 0.6606, + "rewards/accuracies": 0.65625, + "rewards/chosen": -0.08502069115638733, + "rewards/margins": 0.2950421869754791, + "rewards/rejected": -0.38006284832954407, + "step": 818 + }, + { + "epoch": 0.9350723398189723, + "grad_norm": 85.38211010208585, + "learning_rate": 1.2002167934104814e-07, + "logits/chosen": -1.3738641738891602, + "logits/rejected": -1.3845763206481934, + "logps/chosen": -120.14466857910156, + "logps/rejected": -148.59951782226562, + "loss": 0.6025, + "rewards/accuracies": 0.65625, + "rewards/chosen": -0.09485463052988052, + "rewards/margins": 0.2216353416442871, + "rewards/rejected": -0.31648993492126465, + "step": 820 + }, + { + "epoch": 0.9373530040624332, + "grad_norm": 78.59698161262021, + "learning_rate": 1.1964889856260001e-07, + "logits/chosen": -1.405772089958191, + "logits/rejected": -1.4993829727172852, + "logps/chosen": -135.0908203125, + "logps/rejected": -158.73516845703125, + "loss": 0.6449, + "rewards/accuracies": 0.59375, + "rewards/chosen": 0.057033490389585495, + "rewards/margins": 0.19890281558036804, + "rewards/rejected": -0.14186930656433105, + "step": 822 + }, + { + "epoch": 0.9396336683058941, + "grad_norm": 77.09101408818802, + "learning_rate": 1.1927583354929392e-07, + "logits/chosen": -1.2669503688812256, + "logits/rejected": -1.3981753587722778, + "logps/chosen": -185.62281799316406, + "logps/rejected": -229.72349548339844, + "loss": 0.5898, + "rewards/accuracies": 0.78125, + "rewards/chosen": -0.05814466252923012, + "rewards/margins": 0.5641728043556213, + "rewards/rejected": -0.6223174333572388, + "step": 824 + }, + { + "epoch": 0.941914332549355, + "grad_norm": 80.62233528509246, + "learning_rate": 1.1890248969777239e-07, + "logits/chosen": -1.4591572284698486, + "logits/rejected": -1.4886600971221924, + "logps/chosen": -133.4588165283203, + "logps/rejected": -161.96484375, + "loss": 0.6355, + "rewards/accuracies": 0.71875, + "rewards/chosen": -0.1525450050830841, + "rewards/margins": 0.35945460200309753, + "rewards/rejected": -0.5119996666908264, + "step": 826 + }, + { + "epoch": 0.9441949967928159, + "grad_norm": 82.91104665388441, + "learning_rate": 1.1852887240871144e-07, + "logits/chosen": -1.3768235445022583, + "logits/rejected": -1.4797062873840332, + "logps/chosen": -168.19825744628906, + "logps/rejected": -195.83607482910156, + "loss": 0.6328, + "rewards/accuracies": 0.75, + "rewards/chosen": 0.021572627127170563, + "rewards/margins": 0.4793768525123596, + "rewards/rejected": -0.45780429244041443, + "step": 828 + }, + { + "epoch": 0.9464756610362768, + "grad_norm": 99.28432995782116, + "learning_rate": 1.1815498708674265e-07, + "logits/chosen": -1.3641839027404785, + "logits/rejected": -1.5057623386383057, + "logps/chosen": -188.45806884765625, + "logps/rejected": -205.60987854003906, + "loss": 0.6396, + "rewards/accuracies": 0.5, + "rewards/chosen": -0.27040964365005493, + "rewards/margins": 0.16016271710395813, + "rewards/rejected": -0.4305723309516907, + "step": 830 + }, + { + "epoch": 0.9487563252797377, + "grad_norm": 91.59735539302851, + "learning_rate": 1.1778083914037489e-07, + "logits/chosen": -1.4061131477355957, + "logits/rejected": -1.425171136856079, + "logps/chosen": -127.95852661132812, + "logps/rejected": -161.05148315429688, + "loss": 0.6552, + "rewards/accuracies": 0.65625, + "rewards/chosen": 0.016702737659215927, + "rewards/margins": 0.2845785915851593, + "rewards/rejected": -0.2678758502006531, + "step": 832 + }, + { + "epoch": 0.9510369895231986, + "grad_norm": 86.74397487981402, + "learning_rate": 1.17406433981916e-07, + "logits/chosen": -1.3160350322723389, + "logits/rejected": -1.3871701955795288, + "logps/chosen": -176.73638916015625, + "logps/rejected": -211.8169403076172, + "loss": 0.5804, + "rewards/accuracies": 0.75, + "rewards/chosen": -0.43320897221565247, + "rewards/margins": 0.319485068321228, + "rewards/rejected": -0.7526940107345581, + "step": 834 + }, + { + "epoch": 0.9533176537666596, + "grad_norm": 95.1172952955302, + "learning_rate": 1.1703177702739459e-07, + "logits/chosen": -1.4391237497329712, + "logits/rejected": -1.4496339559555054, + "logps/chosen": -170.07687377929688, + "logps/rejected": -187.2700958251953, + "loss": 0.6198, + "rewards/accuracies": 0.59375, + "rewards/chosen": -0.08456218987703323, + "rewards/margins": 0.2940000295639038, + "rewards/rejected": -0.37856221199035645, + "step": 836 + }, + { + "epoch": 0.9555983180101204, + "grad_norm": 88.76345502432684, + "learning_rate": 1.1665687369648172e-07, + "logits/chosen": -1.323454737663269, + "logits/rejected": -1.330315351486206, + "logps/chosen": -124.01346588134766, + "logps/rejected": -141.37887573242188, + "loss": 0.6466, + "rewards/accuracies": 0.5, + "rewards/chosen": -0.2537275552749634, + "rewards/margins": -0.010971667245030403, + "rewards/rejected": -0.24275588989257812, + "step": 838 + }, + { + "epoch": 0.9578789822535814, + "grad_norm": 88.84157969865495, + "learning_rate": 1.1628172941241239e-07, + "logits/chosen": -1.2515794038772583, + "logits/rejected": -1.2858678102493286, + "logps/chosen": -161.89578247070312, + "logps/rejected": -192.5193328857422, + "loss": 0.5897, + "rewards/accuracies": 0.65625, + "rewards/chosen": -0.062142424285411835, + "rewards/margins": 0.38304537534713745, + "rewards/rejected": -0.44518783688545227, + "step": 840 + }, + { + "epoch": 0.9601596464970422, + "grad_norm": 77.39513907682372, + "learning_rate": 1.159063496019072e-07, + "logits/chosen": -1.3906288146972656, + "logits/rejected": -1.529634952545166, + "logps/chosen": -149.05487060546875, + "logps/rejected": -175.32589721679688, + "loss": 0.607, + "rewards/accuracies": 0.46875, + "rewards/chosen": -0.14833964407444, + "rewards/margins": 0.27842655777931213, + "rewards/rejected": -0.42676615715026855, + "step": 842 + }, + { + "epoch": 0.9624403107405032, + "grad_norm": 95.00453640521121, + "learning_rate": 1.1553073969509382e-07, + "logits/chosen": -1.3770264387130737, + "logits/rejected": -1.4574774503707886, + "logps/chosen": -154.23617553710938, + "logps/rejected": -165.87744140625, + "loss": 0.6221, + "rewards/accuracies": 0.65625, + "rewards/chosen": -0.23750467598438263, + "rewards/margins": 0.23638057708740234, + "rewards/rejected": -0.47388529777526855, + "step": 844 + }, + { + "epoch": 0.964720974983964, + "grad_norm": 106.04741673208737, + "learning_rate": 1.1515490512542831e-07, + "logits/chosen": -1.467265248298645, + "logits/rejected": -1.4767037630081177, + "logps/chosen": -154.89556884765625, + "logps/rejected": -169.7006072998047, + "loss": 0.6862, + "rewards/accuracies": 0.6875, + "rewards/chosen": -0.292231023311615, + "rewards/margins": 0.32214629650115967, + "rewards/rejected": -0.6143773198127747, + "step": 846 + }, + { + "epoch": 0.967001639227425, + "grad_norm": 105.71509066302269, + "learning_rate": 1.1477885132961677e-07, + "logits/chosen": -1.3954505920410156, + "logits/rejected": -1.4496076107025146, + "logps/chosen": -135.10621643066406, + "logps/rejected": -155.632080078125, + "loss": 0.6695, + "rewards/accuracies": 0.71875, + "rewards/chosen": -0.11768642067909241, + "rewards/margins": 0.2924567759037018, + "rewards/rejected": -0.4101432263851166, + "step": 848 + }, + { + "epoch": 0.9692823034708858, + "grad_norm": 83.14192239326704, + "learning_rate": 1.1440258374753649e-07, + "logits/chosen": -1.2167763710021973, + "logits/rejected": -1.2719136476516724, + "logps/chosen": -161.74093627929688, + "logps/rejected": -186.17892456054688, + "loss": 0.6215, + "rewards/accuracies": 0.65625, + "rewards/chosen": -0.2962597608566284, + "rewards/margins": 0.19979730248451233, + "rewards/rejected": -0.49605703353881836, + "step": 850 + }, + { + "epoch": 0.9715629677143468, + "grad_norm": 91.00056492305369, + "learning_rate": 1.1402610782215733e-07, + "logits/chosen": -1.338948369026184, + "logits/rejected": -1.4312294721603394, + "logps/chosen": -184.9549560546875, + "logps/rejected": -219.67489624023438, + "loss": 0.6323, + "rewards/accuracies": 0.5625, + "rewards/chosen": -0.3477857708930969, + "rewards/margins": 0.11345378309488297, + "rewards/rejected": -0.46123960614204407, + "step": 852 + }, + { + "epoch": 0.9738436319578078, + "grad_norm": 83.43376640938686, + "learning_rate": 1.1364942899946299e-07, + "logits/chosen": -1.4227657318115234, + "logits/rejected": -1.4756063222885132, + "logps/chosen": -176.06297302246094, + "logps/rejected": -190.72410583496094, + "loss": 0.6332, + "rewards/accuracies": 0.65625, + "rewards/chosen": -0.2725405991077423, + "rewards/margins": 0.23061612248420715, + "rewards/rejected": -0.5031567215919495, + "step": 854 + }, + { + "epoch": 0.9761242962012686, + "grad_norm": 80.98108181650579, + "learning_rate": 1.132725527283722e-07, + "logits/chosen": -1.390743613243103, + "logits/rejected": -1.4082766771316528, + "logps/chosen": -148.54055786132812, + "logps/rejected": -182.93699645996094, + "loss": 0.6133, + "rewards/accuracies": 0.75, + "rewards/chosen": -0.07731722295284271, + "rewards/margins": 0.25885114073753357, + "rewards/rejected": -0.3361683785915375, + "step": 856 + }, + { + "epoch": 0.9784049604447296, + "grad_norm": 84.02405944942352, + "learning_rate": 1.1289548446065993e-07, + "logits/chosen": -1.3017364740371704, + "logits/rejected": -1.3097370862960815, + "logps/chosen": -162.28753662109375, + "logps/rejected": -187.27439880371094, + "loss": 0.6506, + "rewards/accuracies": 0.6875, + "rewards/chosen": -0.13367247581481934, + "rewards/margins": 0.2926619052886963, + "rewards/rejected": -0.426334410905838, + "step": 858 + }, + { + "epoch": 0.9806856246881904, + "grad_norm": 92.76881750126093, + "learning_rate": 1.1251822965087854e-07, + "logits/chosen": -1.2643218040466309, + "logits/rejected": -1.4101805686950684, + "logps/chosen": -162.29969787597656, + "logps/rejected": -204.03713989257812, + "loss": 0.6025, + "rewards/accuracies": 0.6875, + "rewards/chosen": -0.28711238503456116, + "rewards/margins": 0.36070847511291504, + "rewards/rejected": -0.6478208899497986, + "step": 860 + }, + { + "epoch": 0.9829662889316514, + "grad_norm": 118.85388390167415, + "learning_rate": 1.1214079375627883e-07, + "logits/chosen": -1.4487006664276123, + "logits/rejected": -1.474176049232483, + "logps/chosen": -271.73126220703125, + "logps/rejected": -296.800537109375, + "loss": 0.6742, + "rewards/accuracies": 0.59375, + "rewards/chosen": -0.5793529152870178, + "rewards/margins": 0.24643933773040771, + "rewards/rejected": -0.8257922530174255, + "step": 862 + }, + { + "epoch": 0.9852469531751122, + "grad_norm": 86.67252914799577, + "learning_rate": 1.1176318223673105e-07, + "logits/chosen": -1.4337862730026245, + "logits/rejected": -1.4527971744537354, + "logps/chosen": -200.96485900878906, + "logps/rejected": -221.7783660888672, + "loss": 0.6498, + "rewards/accuracies": 0.625, + "rewards/chosen": -0.2869877815246582, + "rewards/margins": 0.2341936230659485, + "rewards/rejected": -0.5211814045906067, + "step": 864 + }, + { + "epoch": 0.9875276174185732, + "grad_norm": 113.98036843821767, + "learning_rate": 1.1138540055464609e-07, + "logits/chosen": -1.3671294450759888, + "logits/rejected": -1.368064522743225, + "logps/chosen": -163.43055725097656, + "logps/rejected": -175.2020721435547, + "loss": 0.5793, + "rewards/accuracies": 0.65625, + "rewards/chosen": -0.022854965180158615, + "rewards/margins": 0.4233308434486389, + "rewards/rejected": -0.44618579745292664, + "step": 866 + }, + { + "epoch": 0.989808281662034, + "grad_norm": 80.29058110520482, + "learning_rate": 1.110074541748963e-07, + "logits/chosen": -1.346540093421936, + "logits/rejected": -1.3472204208374023, + "logps/chosen": -186.08084106445312, + "logps/rejected": -207.60470581054688, + "loss": 0.6321, + "rewards/accuracies": 0.6875, + "rewards/chosen": -0.25492382049560547, + "rewards/margins": 0.35832715034484863, + "rewards/rejected": -0.6132509708404541, + "step": 868 + }, + { + "epoch": 0.992088945905495, + "grad_norm": 90.95115779695014, + "learning_rate": 1.1062934856473653e-07, + "logits/chosen": -1.4387091398239136, + "logits/rejected": -1.455345630645752, + "logps/chosen": -98.49878692626953, + "logps/rejected": -172.24810791015625, + "loss": 0.6444, + "rewards/accuracies": 0.625, + "rewards/chosen": -0.03821319341659546, + "rewards/margins": 0.8005204796791077, + "rewards/rejected": -0.8387336730957031, + "step": 870 + }, + { + "epoch": 0.9943696101489559, + "grad_norm": 100.52505053286734, + "learning_rate": 1.1025108919372499e-07, + "logits/chosen": -1.3714473247528076, + "logits/rejected": -1.3371658325195312, + "logps/chosen": -167.58447265625, + "logps/rejected": -186.04751586914062, + "loss": 0.6675, + "rewards/accuracies": 0.65625, + "rewards/chosen": -0.16858822107315063, + "rewards/margins": 0.19196297228336334, + "rewards/rejected": -0.36055123805999756, + "step": 872 + }, + { + "epoch": 0.9966502743924168, + "grad_norm": 110.94600706114623, + "learning_rate": 1.0987268153364411e-07, + "logits/chosen": -1.3899271488189697, + "logits/rejected": -1.3658580780029297, + "logps/chosen": -130.26220703125, + "logps/rejected": -134.71063232421875, + "loss": 0.66, + "rewards/accuracies": 0.625, + "rewards/chosen": -0.10100418329238892, + "rewards/margins": 0.2500309348106384, + "rewards/rejected": -0.35103511810302734, + "step": 874 + }, + { + "epoch": 0.9989309386358777, + "grad_norm": 105.88720383741304, + "learning_rate": 1.0949413105842146e-07, + "logits/chosen": -1.4607656002044678, + "logits/rejected": -1.5963387489318848, + "logps/chosen": -196.40066528320312, + "logps/rejected": -225.38638305664062, + "loss": 0.6348, + "rewards/accuracies": 0.625, + "rewards/chosen": -0.15517286956310272, + "rewards/margins": 0.15176478028297424, + "rewards/rejected": -0.30693763494491577, + "step": 876 + }, + { + "epoch": 1.0012116028793385, + "grad_norm": 56.3694410581267, + "learning_rate": 1.091154432440506e-07, + "logits/chosen": -1.389705777168274, + "logits/rejected": -1.4425851106643677, + "logps/chosen": -184.08709716796875, + "logps/rejected": -215.4456024169922, + "loss": 0.5323, + "rewards/accuracies": 0.71875, + "rewards/chosen": 0.004449841566383839, + "rewards/margins": 0.635086178779602, + "rewards/rejected": -0.6306363344192505, + "step": 878 + }, + { + "epoch": 1.0034922671227995, + "grad_norm": 59.14871364307595, + "learning_rate": 1.0873662356851164e-07, + "logits/chosen": -1.2264068126678467, + "logits/rejected": -1.2039250135421753, + "logps/chosen": -143.2238311767578, + "logps/rejected": -148.1532745361328, + "loss": 0.4487, + "rewards/accuracies": 0.71875, + "rewards/chosen": 0.14279066026210785, + "rewards/margins": 0.6712049841880798, + "rewards/rejected": -0.528414249420166, + "step": 880 + }, + { + "epoch": 1.0057729313662604, + "grad_norm": 56.08399590075636, + "learning_rate": 1.0835767751169225e-07, + "logits/chosen": -1.3238468170166016, + "logits/rejected": -1.4208531379699707, + "logps/chosen": -172.80247497558594, + "logps/rejected": -221.4132843017578, + "loss": 0.4898, + "rewards/accuracies": 0.75, + "rewards/chosen": -0.04028752073645592, + "rewards/margins": 0.7049790024757385, + "rewards/rejected": -0.7452664375305176, + "step": 882 + }, + { + "epoch": 1.0080535956097214, + "grad_norm": 52.855939708340124, + "learning_rate": 1.0797861055530831e-07, + "logits/chosen": -1.3633054494857788, + "logits/rejected": -1.3413807153701782, + "logps/chosen": -146.97352600097656, + "logps/rejected": -183.86334228515625, + "loss": 0.4712, + "rewards/accuracies": 0.875, + "rewards/chosen": -0.06403888016939163, + "rewards/margins": 0.8986248970031738, + "rewards/rejected": -0.9626636505126953, + "step": 884 + }, + { + "epoch": 1.0103342598531821, + "grad_norm": 66.20782299721449, + "learning_rate": 1.0759942818282453e-07, + "logits/chosen": -1.2454776763916016, + "logits/rejected": -1.3409074544906616, + "logps/chosen": -165.26669311523438, + "logps/rejected": -199.02630615234375, + "loss": 0.4379, + "rewards/accuracies": 0.90625, + "rewards/chosen": -0.18418747186660767, + "rewards/margins": 0.8542241454124451, + "rewards/rejected": -1.0384116172790527, + "step": 886 + }, + { + "epoch": 1.012614924096643, + "grad_norm": 61.538116919835716, + "learning_rate": 1.0722013587937526e-07, + "logits/chosen": -1.345977783203125, + "logits/rejected": -1.3574477434158325, + "logps/chosen": -140.93460083007812, + "logps/rejected": -170.11268615722656, + "loss": 0.4964, + "rewards/accuracies": 0.8125, + "rewards/chosen": 0.001187225803732872, + "rewards/margins": 0.6773791313171387, + "rewards/rejected": -0.6761919260025024, + "step": 888 + }, + { + "epoch": 1.014895588340104, + "grad_norm": 66.18054031071868, + "learning_rate": 1.0684073913168501e-07, + "logits/chosen": -1.506268858909607, + "logits/rejected": -1.5550968647003174, + "logps/chosen": -227.16659545898438, + "logps/rejected": -231.74356079101562, + "loss": 0.4649, + "rewards/accuracies": 0.875, + "rewards/chosen": -0.04133976250886917, + "rewards/margins": 0.7655045986175537, + "rewards/rejected": -0.8068442940711975, + "step": 890 + }, + { + "epoch": 1.017176252583565, + "grad_norm": 63.53563886574213, + "learning_rate": 1.0646124342798919e-07, + "logits/chosen": -1.2807819843292236, + "logits/rejected": -1.3626489639282227, + "logps/chosen": -149.43206787109375, + "logps/rejected": -194.97207641601562, + "loss": 0.4679, + "rewards/accuracies": 0.875, + "rewards/chosen": 0.026193831115961075, + "rewards/margins": 0.7646716237068176, + "rewards/rejected": -0.7384778261184692, + "step": 892 + }, + { + "epoch": 1.0194569168270258, + "grad_norm": 63.988279055249954, + "learning_rate": 1.0608165425795468e-07, + "logits/chosen": -1.5072122812271118, + "logits/rejected": -1.5509088039398193, + "logps/chosen": -185.84625244140625, + "logps/rejected": -213.9545440673828, + "loss": 0.5205, + "rewards/accuracies": 0.84375, + "rewards/chosen": -0.03688034415245056, + "rewards/margins": 0.8168615102767944, + "rewards/rejected": -0.8537418842315674, + "step": 894 + }, + { + "epoch": 1.0217375810704867, + "grad_norm": 58.69101964281229, + "learning_rate": 1.0570197711260038e-07, + "logits/chosen": -1.388396978378296, + "logits/rejected": -1.3820849657058716, + "logps/chosen": -127.9742431640625, + "logps/rejected": -137.78054809570312, + "loss": 0.4494, + "rewards/accuracies": 0.875, + "rewards/chosen": 0.10387776046991348, + "rewards/margins": 0.5781891942024231, + "rewards/rejected": -0.47431135177612305, + "step": 896 + }, + { + "epoch": 1.0240182453139477, + "grad_norm": 74.70102903392454, + "learning_rate": 1.0532221748421786e-07, + "logits/chosen": -1.3617676496505737, + "logits/rejected": -1.4110198020935059, + "logps/chosen": -160.6496124267578, + "logps/rejected": -182.05860900878906, + "loss": 0.4881, + "rewards/accuracies": 0.78125, + "rewards/chosen": -0.21052153408527374, + "rewards/margins": 0.5095985531806946, + "rewards/rejected": -0.7201201319694519, + "step": 898 + }, + { + "epoch": 1.0262989095574087, + "grad_norm": 61.26905190831476, + "learning_rate": 1.0494238086629183e-07, + "logits/chosen": -1.4342904090881348, + "logits/rejected": -1.4664863348007202, + "logps/chosen": -173.26043701171875, + "logps/rejected": -202.07130432128906, + "loss": 0.4934, + "rewards/accuracies": 0.78125, + "rewards/chosen": -0.21440905332565308, + "rewards/margins": 0.7449226379394531, + "rewards/rejected": -0.9593316316604614, + "step": 900 + }, + { + "epoch": 1.0262989095574087, + "eval_logits/chosen": -1.4491240978240967, + "eval_logits/rejected": -1.4348580837249756, + "eval_logps/chosen": -129.59962463378906, + "eval_logps/rejected": -134.80783081054688, + "eval_loss": 0.5797997713088989, + "eval_rewards/accuracies": 0.7200000286102295, + "eval_rewards/chosen": 0.029834765940904617, + "eval_rewards/margins": 0.3043439984321594, + "eval_rewards/rejected": -0.2745092511177063, + "eval_runtime": 20.081, + "eval_samples_per_second": 4.98, + "eval_steps_per_second": 1.245, + "step": 900 } ], "logging_steps": 2,