diff --git "a/trainer_state.json" "b/trainer_state.json" new file mode 100644--- /dev/null +++ "b/trainer_state.json" @@ -0,0 +1,6329 @@ +{ + "best_metric": null, + "best_model_checkpoint": null, + "epoch": 1.0, + "eval_steps": 2000, + "global_step": 4168, + "is_hyper_param_search": false, + "is_local_process_zero": true, + "is_world_process_zero": true, + "log_history": [ + { + "epoch": 0.0002399232245681382, + "grad_norm": 11.290426517034053, + "learning_rate": 1.199040767386091e-09, + "logits/chosen": 1.1020113229751587, + "logits/rejected": 1.165069818496704, + "logps/chosen": -165.85887145996094, + "logps/rejected": -178.0421905517578, + "loss": 0.6931, + "rewards/accuracies": 0.0, + "rewards/chosen": 0.0, + "rewards/margins": 0.0, + "rewards/rejected": 0.0, + "step": 1 + }, + { + "epoch": 0.0023992322456813818, + "grad_norm": 12.642674037441155, + "learning_rate": 1.199040767386091e-08, + "logits/chosen": 0.9179312586784363, + "logits/rejected": 0.8261909484863281, + "logps/chosen": -410.791015625, + "logps/rejected": -328.3950500488281, + "loss": 0.6932, + "rewards/accuracies": 0.4722222089767456, + "rewards/chosen": -0.00014805687533225864, + "rewards/margins": 0.0001381068432237953, + "rewards/rejected": -0.00028616367490030825, + "step": 10 + }, + { + "epoch": 0.0047984644913627635, + "grad_norm": 12.791180740303325, + "learning_rate": 2.398081534772182e-08, + "logits/chosen": 0.9256755709648132, + "logits/rejected": 0.9386581182479858, + "logps/chosen": -278.5252380371094, + "logps/rejected": -242.84823608398438, + "loss": 0.6931, + "rewards/accuracies": 0.4749999940395355, + "rewards/chosen": -0.0002471913758199662, + "rewards/margins": -0.0008500738185830414, + "rewards/rejected": 0.000602882297243923, + "step": 20 + }, + { + "epoch": 0.007197696737044146, + "grad_norm": 11.807415189137759, + "learning_rate": 3.597122302158273e-08, + "logits/chosen": 0.7909670472145081, + "logits/rejected": 0.7317233681678772, + "logps/chosen": -273.0486755371094, + "logps/rejected": -276.9691162109375, + "loss": 0.6935, + "rewards/accuracies": 0.44999998807907104, + "rewards/chosen": -0.00020443140238057822, + "rewards/margins": -0.0006267547723837197, + "rewards/rejected": 0.0004223231808282435, + "step": 30 + }, + { + "epoch": 0.009596928982725527, + "grad_norm": 11.404068374494468, + "learning_rate": 4.796163069544364e-08, + "logits/chosen": 0.8638267517089844, + "logits/rejected": 0.8337821960449219, + "logps/chosen": -277.4988098144531, + "logps/rejected": -265.49273681640625, + "loss": 0.6933, + "rewards/accuracies": 0.5249999761581421, + "rewards/chosen": 0.0007294538663700223, + "rewards/margins": 0.0005479886895045638, + "rewards/rejected": 0.0001814651332097128, + "step": 40 + }, + { + "epoch": 0.01199616122840691, + "grad_norm": 13.063518944634659, + "learning_rate": 5.995203836930455e-08, + "logits/chosen": 0.8905746340751648, + "logits/rejected": 0.8993768692016602, + "logps/chosen": -301.5456848144531, + "logps/rejected": -256.28350830078125, + "loss": 0.6932, + "rewards/accuracies": 0.4000000059604645, + "rewards/chosen": 0.0004059553029946983, + "rewards/margins": -0.0005866002757102251, + "rewards/rejected": 0.0009925555204972625, + "step": 50 + }, + { + "epoch": 0.014395393474088292, + "grad_norm": 12.882875429409548, + "learning_rate": 7.194244604316546e-08, + "logits/chosen": 0.8007913827896118, + "logits/rejected": 0.8331505060195923, + "logps/chosen": -318.5068054199219, + "logps/rejected": -284.26190185546875, + "loss": 0.6931, + "rewards/accuracies": 0.375, + "rewards/chosen": -0.00027945329202339053, + "rewards/margins": -0.0007258501136675477, + "rewards/rejected": 0.0004463967925403267, + "step": 60 + }, + { + "epoch": 0.016794625719769675, + "grad_norm": 11.434777428083084, + "learning_rate": 8.393285371702638e-08, + "logits/chosen": 0.8525362014770508, + "logits/rejected": 0.8576486706733704, + "logps/chosen": -307.0369567871094, + "logps/rejected": -293.033447265625, + "loss": 0.6932, + "rewards/accuracies": 0.4749999940395355, + "rewards/chosen": 0.00017868136637844145, + "rewards/margins": -0.0003002768207807094, + "rewards/rejected": 0.00047895809984765947, + "step": 70 + }, + { + "epoch": 0.019193857965451054, + "grad_norm": 12.302340768468458, + "learning_rate": 9.592326139088728e-08, + "logits/chosen": 0.7650710344314575, + "logits/rejected": 0.9396867752075195, + "logps/chosen": -229.7215118408203, + "logps/rejected": -267.52130126953125, + "loss": 0.6932, + "rewards/accuracies": 0.375, + "rewards/chosen": 0.000520746223628521, + "rewards/margins": 0.00032199383713304996, + "rewards/rejected": 0.00019875231373589486, + "step": 80 + }, + { + "epoch": 0.021593090211132437, + "grad_norm": 12.033324705457542, + "learning_rate": 1.0791366906474819e-07, + "logits/chosen": 0.8275500535964966, + "logits/rejected": 0.7720881700515747, + "logps/chosen": -367.6588439941406, + "logps/rejected": -316.90289306640625, + "loss": 0.6926, + "rewards/accuracies": 0.550000011920929, + "rewards/chosen": -0.0005487919552251697, + "rewards/margins": -0.00021695331088267267, + "rewards/rejected": -0.000331838644342497, + "step": 90 + }, + { + "epoch": 0.02399232245681382, + "grad_norm": 12.472143691246368, + "learning_rate": 1.199040767386091e-07, + "logits/chosen": 0.9897788166999817, + "logits/rejected": 1.1146961450576782, + "logps/chosen": -287.7003173828125, + "logps/rejected": -300.0032958984375, + "loss": 0.6928, + "rewards/accuracies": 0.5249999761581421, + "rewards/chosen": -0.0009165392257273197, + "rewards/margins": 4.667566463467665e-05, + "rewards/rejected": -0.0009632149012759328, + "step": 100 + }, + { + "epoch": 0.026391554702495202, + "grad_norm": 10.47566701912818, + "learning_rate": 1.3189448441247004e-07, + "logits/chosen": 0.862777590751648, + "logits/rejected": 0.8474942445755005, + "logps/chosen": -255.17343139648438, + "logps/rejected": -246.80679321289062, + "loss": 0.6929, + "rewards/accuracies": 0.574999988079071, + "rewards/chosen": -0.0007340327138081193, + "rewards/margins": 9.011449583340436e-05, + "rewards/rejected": -0.0008241472532972693, + "step": 110 + }, + { + "epoch": 0.028790786948176585, + "grad_norm": 11.729418421123865, + "learning_rate": 1.4388489208633092e-07, + "logits/chosen": 0.8798221349716187, + "logits/rejected": 0.8902707099914551, + "logps/chosen": -331.7406311035156, + "logps/rejected": -299.10302734375, + "loss": 0.6926, + "rewards/accuracies": 0.550000011920929, + "rewards/chosen": -0.0016034040600061417, + "rewards/margins": 0.0008947736350819468, + "rewards/rejected": -0.002498177345842123, + "step": 120 + }, + { + "epoch": 0.031190019193857964, + "grad_norm": 10.906048267354812, + "learning_rate": 1.5587529976019183e-07, + "logits/chosen": 0.8031892776489258, + "logits/rejected": 0.8155500292778015, + "logps/chosen": -240.92489624023438, + "logps/rejected": -324.6410217285156, + "loss": 0.6921, + "rewards/accuracies": 0.574999988079071, + "rewards/chosen": -0.0009776324732229114, + "rewards/margins": 0.0024681445211172104, + "rewards/rejected": -0.0034457766450941563, + "step": 130 + }, + { + "epoch": 0.03358925143953935, + "grad_norm": 12.066753821066918, + "learning_rate": 1.6786570743405277e-07, + "logits/chosen": 0.9222855567932129, + "logits/rejected": 0.8834908604621887, + "logps/chosen": -314.27435302734375, + "logps/rejected": -301.2237854003906, + "loss": 0.692, + "rewards/accuracies": 0.5249999761581421, + "rewards/chosen": -0.003638929221779108, + "rewards/margins": 0.0016478378092870116, + "rewards/rejected": -0.0052867671474814415, + "step": 140 + }, + { + "epoch": 0.03598848368522073, + "grad_norm": 12.461323392055261, + "learning_rate": 1.7985611510791365e-07, + "logits/chosen": 0.8271953463554382, + "logits/rejected": 0.8963567614555359, + "logps/chosen": -255.406005859375, + "logps/rejected": -246.0330352783203, + "loss": 0.6921, + "rewards/accuracies": 0.6000000238418579, + "rewards/chosen": -0.003589408006519079, + "rewards/margins": 0.003603331744670868, + "rewards/rejected": -0.007192739751189947, + "step": 150 + }, + { + "epoch": 0.03838771593090211, + "grad_norm": 12.837513374911397, + "learning_rate": 1.9184652278177456e-07, + "logits/chosen": 0.8090554475784302, + "logits/rejected": 0.7438990473747253, + "logps/chosen": -327.182373046875, + "logps/rejected": -251.8794708251953, + "loss": 0.6905, + "rewards/accuracies": 0.6000000238418579, + "rewards/chosen": -0.003899016883224249, + "rewards/margins": 0.0035891197621822357, + "rewards/rejected": -0.007488137576729059, + "step": 160 + }, + { + "epoch": 0.040786948176583494, + "grad_norm": 11.693960987426166, + "learning_rate": 2.038369304556355e-07, + "logits/chosen": 0.9418096542358398, + "logits/rejected": 0.9179880023002625, + "logps/chosen": -373.8608093261719, + "logps/rejected": -364.60565185546875, + "loss": 0.6906, + "rewards/accuracies": 0.5249999761581421, + "rewards/chosen": -0.008983137086033821, + "rewards/margins": 0.0037296731024980545, + "rewards/rejected": -0.012712809257209301, + "step": 170 + }, + { + "epoch": 0.04318618042226487, + "grad_norm": 13.093807399181138, + "learning_rate": 2.1582733812949638e-07, + "logits/chosen": 0.7935225367546082, + "logits/rejected": 0.8044607043266296, + "logps/chosen": -255.5353546142578, + "logps/rejected": -246.3568115234375, + "loss": 0.6909, + "rewards/accuracies": 0.550000011920929, + "rewards/chosen": -0.006067480426281691, + "rewards/margins": 0.00562057550996542, + "rewards/rejected": -0.011688055470585823, + "step": 180 + }, + { + "epoch": 0.04558541266794626, + "grad_norm": 14.283584623097973, + "learning_rate": 2.278177458033573e-07, + "logits/chosen": 0.8921982645988464, + "logits/rejected": 0.875672459602356, + "logps/chosen": -337.9564514160156, + "logps/rejected": -272.2162170410156, + "loss": 0.6892, + "rewards/accuracies": 0.675000011920929, + "rewards/chosen": -0.008288905955851078, + "rewards/margins": 0.007777436636388302, + "rewards/rejected": -0.01606634259223938, + "step": 190 + }, + { + "epoch": 0.04798464491362764, + "grad_norm": 10.889194335114492, + "learning_rate": 2.398081534772182e-07, + "logits/chosen": 0.7922821640968323, + "logits/rejected": 0.8541098833084106, + "logps/chosen": -336.982177734375, + "logps/rejected": -318.88372802734375, + "loss": 0.6881, + "rewards/accuracies": 0.7749999761581421, + "rewards/chosen": -0.008277793414890766, + "rewards/margins": 0.013390962965786457, + "rewards/rejected": -0.021668758243322372, + "step": 200 + }, + { + "epoch": 0.05038387715930902, + "grad_norm": 12.286406041336088, + "learning_rate": 2.517985611510791e-07, + "logits/chosen": 0.7591502070426941, + "logits/rejected": 0.7965251803398132, + "logps/chosen": -256.0464172363281, + "logps/rejected": -279.0765380859375, + "loss": 0.6869, + "rewards/accuracies": 0.7250000238418579, + "rewards/chosen": -0.015102909877896309, + "rewards/margins": 0.008513028733432293, + "rewards/rejected": -0.023615941405296326, + "step": 210 + }, + { + "epoch": 0.052783109404990404, + "grad_norm": 11.650837722032147, + "learning_rate": 2.637889688249401e-07, + "logits/chosen": 0.8363003730773926, + "logits/rejected": 0.8418310880661011, + "logps/chosen": -341.4637756347656, + "logps/rejected": -334.71722412109375, + "loss": 0.6873, + "rewards/accuracies": 0.550000011920929, + "rewards/chosen": -0.021802758798003197, + "rewards/margins": 0.012052404694259167, + "rewards/rejected": -0.03385516256093979, + "step": 220 + }, + { + "epoch": 0.05518234165067178, + "grad_norm": 12.17094186601859, + "learning_rate": 2.7577937649880093e-07, + "logits/chosen": 0.9586321711540222, + "logits/rejected": 1.0120365619659424, + "logps/chosen": -258.8882141113281, + "logps/rejected": -295.0778503417969, + "loss": 0.6829, + "rewards/accuracies": 0.625, + "rewards/chosen": -0.021029580384492874, + "rewards/margins": 0.024138668552041054, + "rewards/rejected": -0.04516825079917908, + "step": 230 + }, + { + "epoch": 0.05758157389635317, + "grad_norm": 13.555388385039116, + "learning_rate": 2.8776978417266184e-07, + "logits/chosen": 0.8616135716438293, + "logits/rejected": 0.860598087310791, + "logps/chosen": -327.2540588378906, + "logps/rejected": -277.73858642578125, + "loss": 0.6805, + "rewards/accuracies": 0.7749999761581421, + "rewards/chosen": -0.025776376947760582, + "rewards/margins": 0.029360895976424217, + "rewards/rejected": -0.0551372691988945, + "step": 240 + }, + { + "epoch": 0.05998080614203455, + "grad_norm": 12.974446675999463, + "learning_rate": 2.997601918465228e-07, + "logits/chosen": 0.9154967069625854, + "logits/rejected": 0.953376293182373, + "logps/chosen": -261.62432861328125, + "logps/rejected": -251.11181640625, + "loss": 0.6805, + "rewards/accuracies": 0.6499999761581421, + "rewards/chosen": -0.035514116287231445, + "rewards/margins": 0.020228754729032516, + "rewards/rejected": -0.05574287101626396, + "step": 250 + }, + { + "epoch": 0.06238003838771593, + "grad_norm": 12.083196573389626, + "learning_rate": 3.1175059952038366e-07, + "logits/chosen": 0.7959989309310913, + "logits/rejected": 0.9321529269218445, + "logps/chosen": -291.46197509765625, + "logps/rejected": -285.123291015625, + "loss": 0.6749, + "rewards/accuracies": 0.7749999761581421, + "rewards/chosen": -0.039493151009082794, + "rewards/margins": 0.040632762014865875, + "rewards/rejected": -0.08012591302394867, + "step": 260 + }, + { + "epoch": 0.0647792706333973, + "grad_norm": 12.75510090459433, + "learning_rate": 3.2374100719424457e-07, + "logits/chosen": 0.9645400047302246, + "logits/rejected": 0.8777379989624023, + "logps/chosen": -312.07080078125, + "logps/rejected": -250.50857543945312, + "loss": 0.6719, + "rewards/accuracies": 0.6000000238418579, + "rewards/chosen": -0.04036339372396469, + "rewards/margins": 0.02411934733390808, + "rewards/rejected": -0.06448274105787277, + "step": 270 + }, + { + "epoch": 0.0671785028790787, + "grad_norm": 13.510481134396441, + "learning_rate": 3.3573141486810554e-07, + "logits/chosen": 0.8545993566513062, + "logits/rejected": 0.9058882594108582, + "logps/chosen": -325.89508056640625, + "logps/rejected": -312.71478271484375, + "loss": 0.6621, + "rewards/accuracies": 0.75, + "rewards/chosen": -0.06212998181581497, + "rewards/margins": 0.05864911153912544, + "rewards/rejected": -0.12077908217906952, + "step": 280 + }, + { + "epoch": 0.06957773512476008, + "grad_norm": 11.640828173777832, + "learning_rate": 3.477218225419664e-07, + "logits/chosen": 0.9046756029129028, + "logits/rejected": 0.9748743176460266, + "logps/chosen": -317.8568420410156, + "logps/rejected": -289.9772033691406, + "loss": 0.6637, + "rewards/accuracies": 0.800000011920929, + "rewards/chosen": -0.08539162576198578, + "rewards/margins": 0.06998683512210846, + "rewards/rejected": -0.15537847578525543, + "step": 290 + }, + { + "epoch": 0.07197696737044146, + "grad_norm": 13.967496487157206, + "learning_rate": 3.597122302158273e-07, + "logits/chosen": 0.8389209508895874, + "logits/rejected": 0.7966829538345337, + "logps/chosen": -292.8455505371094, + "logps/rejected": -308.762451171875, + "loss": 0.6617, + "rewards/accuracies": 0.6000000238418579, + "rewards/chosen": -0.12145288288593292, + "rewards/margins": 0.07226552069187164, + "rewards/rejected": -0.19371840357780457, + "step": 300 + }, + { + "epoch": 0.07437619961612284, + "grad_norm": 11.85245858072109, + "learning_rate": 3.7170263788968827e-07, + "logits/chosen": 0.9054175615310669, + "logits/rejected": 0.8772112727165222, + "logps/chosen": -306.7538146972656, + "logps/rejected": -262.8247985839844, + "loss": 0.6658, + "rewards/accuracies": 0.7250000238418579, + "rewards/chosen": -0.13142530620098114, + "rewards/margins": 0.0997721403837204, + "rewards/rejected": -0.23119744658470154, + "step": 310 + }, + { + "epoch": 0.07677543186180422, + "grad_norm": 11.983652317145598, + "learning_rate": 3.836930455635491e-07, + "logits/chosen": 0.8777133226394653, + "logits/rejected": 0.880419909954071, + "logps/chosen": -313.6241455078125, + "logps/rejected": -279.3583984375, + "loss": 0.6609, + "rewards/accuracies": 0.625, + "rewards/chosen": -0.1642109900712967, + "rewards/margins": 0.06335286051034927, + "rewards/rejected": -0.22756382822990417, + "step": 320 + }, + { + "epoch": 0.07917466410748561, + "grad_norm": 12.597865260643777, + "learning_rate": 3.9568345323741003e-07, + "logits/chosen": 0.8802727460861206, + "logits/rejected": 0.9690290689468384, + "logps/chosen": -291.7769470214844, + "logps/rejected": -340.9366455078125, + "loss": 0.6498, + "rewards/accuracies": 0.625, + "rewards/chosen": -0.2159583568572998, + "rewards/margins": 0.12743426859378815, + "rewards/rejected": -0.34339264035224915, + "step": 330 + }, + { + "epoch": 0.08157389635316699, + "grad_norm": 12.031574034409454, + "learning_rate": 4.07673860911271e-07, + "logits/chosen": 1.0568187236785889, + "logits/rejected": 0.9909990429878235, + "logps/chosen": -276.7728576660156, + "logps/rejected": -308.56048583984375, + "loss": 0.6486, + "rewards/accuracies": 0.75, + "rewards/chosen": -0.18150362372398376, + "rewards/margins": 0.15542708337306976, + "rewards/rejected": -0.33693069219589233, + "step": 340 + }, + { + "epoch": 0.08397312859884837, + "grad_norm": 14.09254079956678, + "learning_rate": 4.1966426858513185e-07, + "logits/chosen": 0.7211519479751587, + "logits/rejected": 0.715661883354187, + "logps/chosen": -328.1579284667969, + "logps/rejected": -335.1773986816406, + "loss": 0.6547, + "rewards/accuracies": 0.6499999761581421, + "rewards/chosen": -0.2924322783946991, + "rewards/margins": 0.11239685118198395, + "rewards/rejected": -0.40482911467552185, + "step": 350 + }, + { + "epoch": 0.08637236084452975, + "grad_norm": 14.373438087738135, + "learning_rate": 4.3165467625899276e-07, + "logits/chosen": 1.0743309259414673, + "logits/rejected": 0.9853240251541138, + "logps/chosen": -320.19158935546875, + "logps/rejected": -270.3163146972656, + "loss": 0.6509, + "rewards/accuracies": 0.699999988079071, + "rewards/chosen": -0.3367859423160553, + "rewards/margins": 0.08087924122810364, + "rewards/rejected": -0.41766518354415894, + "step": 360 + }, + { + "epoch": 0.08877159309021113, + "grad_norm": 14.669348649143826, + "learning_rate": 4.436450839328537e-07, + "logits/chosen": 0.9024693369865417, + "logits/rejected": 1.007645606994629, + "logps/chosen": -297.4977111816406, + "logps/rejected": -319.9523010253906, + "loss": 0.6372, + "rewards/accuracies": 0.7749999761581421, + "rewards/chosen": -0.3957563638687134, + "rewards/margins": 0.1617899388074875, + "rewards/rejected": -0.5575462579727173, + "step": 370 + }, + { + "epoch": 0.09117082533589252, + "grad_norm": 12.430586958233993, + "learning_rate": 4.556354916067146e-07, + "logits/chosen": 0.7547154426574707, + "logits/rejected": 0.9096601605415344, + "logps/chosen": -291.98260498046875, + "logps/rejected": -299.932861328125, + "loss": 0.618, + "rewards/accuracies": 0.75, + "rewards/chosen": -0.26660218834877014, + "rewards/margins": 0.16147901117801666, + "rewards/rejected": -0.42808112502098083, + "step": 380 + }, + { + "epoch": 0.0935700575815739, + "grad_norm": 15.743603343827022, + "learning_rate": 4.676258992805755e-07, + "logits/chosen": 0.9608089327812195, + "logits/rejected": 1.006642460823059, + "logps/chosen": -337.75048828125, + "logps/rejected": -310.911376953125, + "loss": 0.6329, + "rewards/accuracies": 0.75, + "rewards/chosen": -0.4503099322319031, + "rewards/margins": 0.10561531782150269, + "rewards/rejected": -0.5559252500534058, + "step": 390 + }, + { + "epoch": 0.09596928982725528, + "grad_norm": 13.296795216212011, + "learning_rate": 4.796163069544364e-07, + "logits/chosen": 0.9124090075492859, + "logits/rejected": 0.9337595701217651, + "logps/chosen": -335.5029296875, + "logps/rejected": -336.509033203125, + "loss": 0.6271, + "rewards/accuracies": 0.6499999761581421, + "rewards/chosen": -0.5172705054283142, + "rewards/margins": 0.24937620759010315, + "rewards/rejected": -0.7666467428207397, + "step": 400 + }, + { + "epoch": 0.09836852207293666, + "grad_norm": 16.087465846751012, + "learning_rate": 4.916067146282974e-07, + "logits/chosen": 0.77690190076828, + "logits/rejected": 0.8211329579353333, + "logps/chosen": -319.15301513671875, + "logps/rejected": -365.4151611328125, + "loss": 0.6109, + "rewards/accuracies": 0.699999988079071, + "rewards/chosen": -0.5081580877304077, + "rewards/margins": 0.23018617928028107, + "rewards/rejected": -0.7383443117141724, + "step": 410 + }, + { + "epoch": 0.10076775431861804, + "grad_norm": 17.466087793122092, + "learning_rate": 4.999992108529978e-07, + "logits/chosen": 0.9634016752243042, + "logits/rejected": 1.0442949533462524, + "logps/chosen": -432.4740295410156, + "logps/rejected": -414.32666015625, + "loss": 0.6331, + "rewards/accuracies": 0.699999988079071, + "rewards/chosen": -0.6796163320541382, + "rewards/margins": 0.297426700592041, + "rewards/rejected": -0.9770429730415344, + "step": 420 + }, + { + "epoch": 0.10316698656429943, + "grad_norm": 17.06585616280862, + "learning_rate": 4.999851817115532e-07, + "logits/chosen": 0.8025485277175903, + "logits/rejected": 0.9477583169937134, + "logps/chosen": -335.64129638671875, + "logps/rejected": -348.13201904296875, + "loss": 0.6349, + "rewards/accuracies": 0.550000011920929, + "rewards/chosen": -0.5576916933059692, + "rewards/margins": 0.2742641270160675, + "rewards/rejected": -0.8319557905197144, + "step": 430 + }, + { + "epoch": 0.10556621880998081, + "grad_norm": 14.864309504803636, + "learning_rate": 4.999536171027889e-07, + "logits/chosen": 0.8711138963699341, + "logits/rejected": 0.8473278880119324, + "logps/chosen": -354.7337646484375, + "logps/rejected": -355.3828125, + "loss": 0.6349, + "rewards/accuracies": 0.699999988079071, + "rewards/chosen": -0.6624523401260376, + "rewards/margins": 0.12896910309791565, + "rewards/rejected": -0.7914214134216309, + "step": 440 + }, + { + "epoch": 0.10796545105566219, + "grad_norm": 14.807426790938989, + "learning_rate": 4.999045192408369e-07, + "logits/chosen": 0.7770389318466187, + "logits/rejected": 0.7765820622444153, + "logps/chosen": -323.10968017578125, + "logps/rejected": -317.7047119140625, + "loss": 0.6344, + "rewards/accuracies": 0.6000000238418579, + "rewards/chosen": -0.6039828062057495, + "rewards/margins": 0.15776701271533966, + "rewards/rejected": -0.761749804019928, + "step": 450 + }, + { + "epoch": 0.11036468330134357, + "grad_norm": 15.367304865931045, + "learning_rate": 4.998378915697171e-07, + "logits/chosen": 0.880317211151123, + "logits/rejected": 0.842047393321991, + "logps/chosen": -346.06866455078125, + "logps/rejected": -363.48333740234375, + "loss": 0.6047, + "rewards/accuracies": 0.699999988079071, + "rewards/chosen": -0.40573984384536743, + "rewards/margins": 0.29732540249824524, + "rewards/rejected": -0.7030652165412903, + "step": 460 + }, + { + "epoch": 0.11276391554702495, + "grad_norm": 15.949013687358022, + "learning_rate": 4.997537387630958e-07, + "logits/chosen": 0.7507516741752625, + "logits/rejected": 0.8210875391960144, + "logps/chosen": -288.9540710449219, + "logps/rejected": -307.3976135253906, + "loss": 0.602, + "rewards/accuracies": 0.6000000238418579, + "rewards/chosen": -0.48658061027526855, + "rewards/margins": 0.1853361576795578, + "rewards/rejected": -0.671916663646698, + "step": 470 + }, + { + "epoch": 0.11516314779270634, + "grad_norm": 16.9694155509272, + "learning_rate": 4.996520667239582e-07, + "logits/chosen": 0.7397254109382629, + "logits/rejected": 0.9247487187385559, + "logps/chosen": -316.98785400390625, + "logps/rejected": -398.54302978515625, + "loss": 0.5981, + "rewards/accuracies": 0.675000011920929, + "rewards/chosen": -0.48844990134239197, + "rewards/margins": 0.33046063780784607, + "rewards/rejected": -0.8189105987548828, + "step": 480 + }, + { + "epoch": 0.11756238003838772, + "grad_norm": 15.78316684748212, + "learning_rate": 4.995328825841939e-07, + "logits/chosen": 0.8742391467094421, + "logits/rejected": 0.8716489672660828, + "logps/chosen": -293.3537292480469, + "logps/rejected": -325.4706726074219, + "loss": 0.5906, + "rewards/accuracies": 0.7749999761581421, + "rewards/chosen": -0.3982633650302887, + "rewards/margins": 0.4348866045475006, + "rewards/rejected": -0.8331500291824341, + "step": 490 + }, + { + "epoch": 0.1199616122840691, + "grad_norm": 17.430261021826254, + "learning_rate": 4.993961947040967e-07, + "logits/chosen": 0.8817283511161804, + "logits/rejected": 0.8894240260124207, + "logps/chosen": -379.4447326660156, + "logps/rejected": -356.8714599609375, + "loss": 0.6058, + "rewards/accuracies": 0.699999988079071, + "rewards/chosen": -0.7444159388542175, + "rewards/margins": 0.18644893169403076, + "rewards/rejected": -0.9308649301528931, + "step": 500 + }, + { + "epoch": 0.12236084452975048, + "grad_norm": 15.352342487594235, + "learning_rate": 4.992420126717784e-07, + "logits/chosen": 0.7498997449874878, + "logits/rejected": 0.8734132647514343, + "logps/chosen": -334.142822265625, + "logps/rejected": -378.78839111328125, + "loss": 0.608, + "rewards/accuracies": 0.875, + "rewards/chosen": -0.5240365862846375, + "rewards/margins": 0.5625481605529785, + "rewards/rejected": -1.0865848064422607, + "step": 510 + }, + { + "epoch": 0.12476007677543186, + "grad_norm": 17.82749679734051, + "learning_rate": 4.990703473024958e-07, + "logits/chosen": 0.7966002225875854, + "logits/rejected": 0.7500957250595093, + "logps/chosen": -385.6920166015625, + "logps/rejected": -407.591552734375, + "loss": 0.6209, + "rewards/accuracies": 0.675000011920929, + "rewards/chosen": -0.7459064722061157, + "rewards/margins": 0.31928759813308716, + "rewards/rejected": -1.0651941299438477, + "step": 520 + }, + { + "epoch": 0.12715930902111325, + "grad_norm": 16.784735945624263, + "learning_rate": 4.98881210637893e-07, + "logits/chosen": 0.6152772903442383, + "logits/rejected": 0.7151674032211304, + "logps/chosen": -299.0379943847656, + "logps/rejected": -367.5239562988281, + "loss": 0.6181, + "rewards/accuracies": 0.7749999761581421, + "rewards/chosen": -0.5808143615722656, + "rewards/margins": 0.37133264541625977, + "rewards/rejected": -0.9521470069885254, + "step": 530 + }, + { + "epoch": 0.1295585412667946, + "grad_norm": 15.174838695691413, + "learning_rate": 4.986746159451553e-07, + "logits/chosen": 0.6313377618789673, + "logits/rejected": 0.6896692514419556, + "logps/chosen": -331.22296142578125, + "logps/rejected": -373.2359313964844, + "loss": 0.5974, + "rewards/accuracies": 0.75, + "rewards/chosen": -0.5576273798942566, + "rewards/margins": 0.5619553327560425, + "rewards/rejected": -1.1195827722549438, + "step": 540 + }, + { + "epoch": 0.131957773512476, + "grad_norm": 14.111267602676048, + "learning_rate": 4.984505777160795e-07, + "logits/chosen": 0.7313958406448364, + "logits/rejected": 0.6984618902206421, + "logps/chosen": -400.5677795410156, + "logps/rejected": -427.5818786621094, + "loss": 0.6113, + "rewards/accuracies": 0.675000011920929, + "rewards/chosen": -0.7098091840744019, + "rewards/margins": 0.32652807235717773, + "rewards/rejected": -1.03633713722229, + "step": 550 + }, + { + "epoch": 0.1343570057581574, + "grad_norm": 16.422136951074723, + "learning_rate": 4.982091116660574e-07, + "logits/chosen": 0.7890074253082275, + "logits/rejected": 0.697285532951355, + "logps/chosen": -271.3017272949219, + "logps/rejected": -260.1424865722656, + "loss": 0.628, + "rewards/accuracies": 0.550000011920929, + "rewards/chosen": -0.5347421169281006, + "rewards/margins": 0.14706984162330627, + "rewards/rejected": -0.6818119287490845, + "step": 560 + }, + { + "epoch": 0.13675623800383876, + "grad_norm": 18.921421716557443, + "learning_rate": 4.979502347329732e-07, + "logits/chosen": 0.6331374049186707, + "logits/rejected": 0.7241517305374146, + "logps/chosen": -383.59393310546875, + "logps/rejected": -467.81591796875, + "loss": 0.6047, + "rewards/accuracies": 0.699999988079071, + "rewards/chosen": -0.7832542061805725, + "rewards/margins": 0.5469503402709961, + "rewards/rejected": -1.330204725265503, + "step": 570 + }, + { + "epoch": 0.13915547024952016, + "grad_norm": 22.144464177364856, + "learning_rate": 4.976739650760151e-07, + "logits/chosen": 0.6505743265151978, + "logits/rejected": 0.7051428556442261, + "logps/chosen": -349.1499938964844, + "logps/rejected": -378.1747131347656, + "loss": 0.5913, + "rewards/accuracies": 0.6000000238418579, + "rewards/chosen": -0.6620784401893616, + "rewards/margins": 0.4197353422641754, + "rewards/rejected": -1.0818138122558594, + "step": 580 + }, + { + "epoch": 0.14155470249520152, + "grad_norm": 21.853395092107863, + "learning_rate": 4.97380322074402e-07, + "logits/chosen": 0.7972511053085327, + "logits/rejected": 0.7657841444015503, + "logps/chosen": -327.83294677734375, + "logps/rejected": -362.2701721191406, + "loss": 0.6311, + "rewards/accuracies": 0.6000000238418579, + "rewards/chosen": -0.864510715007782, + "rewards/margins": 0.39511948823928833, + "rewards/rejected": -1.2596302032470703, + "step": 590 + }, + { + "epoch": 0.14395393474088292, + "grad_norm": 17.131174415142638, + "learning_rate": 4.970693263260237e-07, + "logits/chosen": 0.7822259068489075, + "logits/rejected": 0.7136083841323853, + "logps/chosen": -390.30084228515625, + "logps/rejected": -380.1099548339844, + "loss": 0.5966, + "rewards/accuracies": 0.7250000238418579, + "rewards/chosen": -0.6453007459640503, + "rewards/margins": 0.44277459383010864, + "rewards/rejected": -1.0880753993988037, + "step": 600 + }, + { + "epoch": 0.1463531669865643, + "grad_norm": 17.663633787684425, + "learning_rate": 4.967409996459966e-07, + "logits/chosen": 0.5881372690200806, + "logits/rejected": 0.6206382513046265, + "logps/chosen": -356.79901123046875, + "logps/rejected": -371.9761047363281, + "loss": 0.5787, + "rewards/accuracies": 0.800000011920929, + "rewards/chosen": -0.6707444787025452, + "rewards/margins": 0.4131700396537781, + "rewards/rejected": -1.0839145183563232, + "step": 610 + }, + { + "epoch": 0.14875239923224567, + "grad_norm": 18.054692834695842, + "learning_rate": 4.963953650651326e-07, + "logits/chosen": 0.8582113981246948, + "logits/rejected": 0.7811607122421265, + "logps/chosen": -460.695556640625, + "logps/rejected": -398.85345458984375, + "loss": 0.6048, + "rewards/accuracies": 0.6499999761581421, + "rewards/chosen": -0.8490235209465027, + "rewards/margins": 0.4152224659919739, + "rewards/rejected": -1.2642459869384766, + "step": 620 + }, + { + "epoch": 0.15115163147792707, + "grad_norm": 16.657440247472014, + "learning_rate": 4.960324468283248e-07, + "logits/chosen": 0.7700496912002563, + "logits/rejected": 0.8080910444259644, + "logps/chosen": -296.64111328125, + "logps/rejected": -339.17388916015625, + "loss": 0.5599, + "rewards/accuracies": 0.7749999761581421, + "rewards/chosen": -0.550514817237854, + "rewards/margins": 0.4359719753265381, + "rewards/rejected": -0.9864869117736816, + "step": 630 + }, + { + "epoch": 0.15355086372360843, + "grad_norm": 18.818954073381697, + "learning_rate": 4.956522703928451e-07, + "logits/chosen": 0.7167537212371826, + "logits/rejected": 0.7465007305145264, + "logps/chosen": -320.0273132324219, + "logps/rejected": -379.2918701171875, + "loss": 0.5728, + "rewards/accuracies": 0.699999988079071, + "rewards/chosen": -0.6476621627807617, + "rewards/margins": 0.5033532977104187, + "rewards/rejected": -1.1510154008865356, + "step": 640 + }, + { + "epoch": 0.15595009596928983, + "grad_norm": 19.432326737759745, + "learning_rate": 4.952548624265606e-07, + "logits/chosen": 0.685808002948761, + "logits/rejected": 0.748884916305542, + "logps/chosen": -381.24188232421875, + "logps/rejected": -390.5279846191406, + "loss": 0.6188, + "rewards/accuracies": 0.6499999761581421, + "rewards/chosen": -0.847141444683075, + "rewards/margins": 0.24964895844459534, + "rewards/rejected": -1.0967904329299927, + "step": 650 + }, + { + "epoch": 0.15834932821497122, + "grad_norm": 15.565167912174749, + "learning_rate": 4.948402508060607e-07, + "logits/chosen": 0.579333484172821, + "logits/rejected": 0.6070585250854492, + "logps/chosen": -319.1553955078125, + "logps/rejected": -362.86920166015625, + "loss": 0.612, + "rewards/accuracies": 0.6499999761581421, + "rewards/chosen": -0.6768738627433777, + "rewards/margins": 0.6027628779411316, + "rewards/rejected": -1.2796367406845093, + "step": 660 + }, + { + "epoch": 0.16074856046065258, + "grad_norm": 19.32420588879935, + "learning_rate": 4.944084646147038e-07, + "logits/chosen": 0.5981010794639587, + "logits/rejected": 0.626388669013977, + "logps/chosen": -407.26708984375, + "logps/rejected": -400.2440490722656, + "loss": 0.6386, + "rewards/accuracies": 0.675000011920929, + "rewards/chosen": -0.66740882396698, + "rewards/margins": 0.18918684124946594, + "rewards/rejected": -0.8565956354141235, + "step": 670 + }, + { + "epoch": 0.16314779270633398, + "grad_norm": 15.975618010502641, + "learning_rate": 4.939595341405754e-07, + "logits/chosen": 0.6756628751754761, + "logits/rejected": 0.6775110960006714, + "logps/chosen": -328.9629821777344, + "logps/rejected": -333.85150146484375, + "loss": 0.6117, + "rewards/accuracies": 0.699999988079071, + "rewards/chosen": -0.5073133111000061, + "rewards/margins": 0.3033245801925659, + "rewards/rejected": -0.8106378316879272, + "step": 680 + }, + { + "epoch": 0.16554702495201534, + "grad_norm": 15.691147213625676, + "learning_rate": 4.93493490874365e-07, + "logits/chosen": 0.7397656440734863, + "logits/rejected": 0.7417260408401489, + "logps/chosen": -319.7871398925781, + "logps/rejected": -361.149658203125, + "loss": 0.5674, + "rewards/accuracies": 0.7250000238418579, + "rewards/chosen": -0.5759695172309875, + "rewards/margins": 0.34965625405311584, + "rewards/rejected": -0.9256256818771362, + "step": 690 + }, + { + "epoch": 0.16794625719769674, + "grad_norm": 16.01928013375132, + "learning_rate": 4.93010367507156e-07, + "logits/chosen": 0.6512678861618042, + "logits/rejected": 0.6891155242919922, + "logps/chosen": -295.92822265625, + "logps/rejected": -338.79608154296875, + "loss": 0.5706, + "rewards/accuracies": 0.6499999761581421, + "rewards/chosen": -0.5925602316856384, + "rewards/margins": 0.686859130859375, + "rewards/rejected": -1.2794193029403687, + "step": 700 + }, + { + "epoch": 0.17034548944337813, + "grad_norm": 19.518133270314312, + "learning_rate": 4.925101979281332e-07, + "logits/chosen": 0.5287461876869202, + "logits/rejected": 0.5766854286193848, + "logps/chosen": -400.12713623046875, + "logps/rejected": -398.5772705078125, + "loss": 0.5982, + "rewards/accuracies": 0.75, + "rewards/chosen": -0.6956884264945984, + "rewards/margins": 0.593571662902832, + "rewards/rejected": -1.2892601490020752, + "step": 710 + }, + { + "epoch": 0.1727447216890595, + "grad_norm": 17.593931676516338, + "learning_rate": 4.919930172222054e-07, + "logits/chosen": 0.5889393091201782, + "logits/rejected": 0.5955343842506409, + "logps/chosen": -341.21307373046875, + "logps/rejected": -389.63897705078125, + "loss": 0.5556, + "rewards/accuracies": 0.800000011920929, + "rewards/chosen": -0.6909217238426208, + "rewards/margins": 0.5461858510971069, + "rewards/rejected": -1.237107515335083, + "step": 720 + }, + { + "epoch": 0.1751439539347409, + "grad_norm": 23.194816745121578, + "learning_rate": 4.914588616675445e-07, + "logits/chosen": 0.48078519105911255, + "logits/rejected": 0.45681601762771606, + "logps/chosen": -333.8201904296875, + "logps/rejected": -357.1300964355469, + "loss": 0.616, + "rewards/accuracies": 0.75, + "rewards/chosen": -0.6982260346412659, + "rewards/margins": 0.46993082761764526, + "rewards/rejected": -1.1681568622589111, + "step": 730 + }, + { + "epoch": 0.17754318618042225, + "grad_norm": 18.34717096478732, + "learning_rate": 4.909077687330404e-07, + "logits/chosen": 0.5538076162338257, + "logits/rejected": 0.596636176109314, + "logps/chosen": -352.73931884765625, + "logps/rejected": -347.88494873046875, + "loss": 0.5689, + "rewards/accuracies": 0.800000011920929, + "rewards/chosen": -0.5955497622489929, + "rewards/margins": 0.2640233635902405, + "rewards/rejected": -0.8595730662345886, + "step": 740 + }, + { + "epoch": 0.17994241842610365, + "grad_norm": 18.33581257388434, + "learning_rate": 4.903397770756729e-07, + "logits/chosen": 0.43998318910598755, + "logits/rejected": 0.47118696570396423, + "logps/chosen": -346.0210266113281, + "logps/rejected": -386.53839111328125, + "loss": 0.5761, + "rewards/accuracies": 0.800000011920929, + "rewards/chosen": -0.6450529098510742, + "rewards/margins": 0.45645374059677124, + "rewards/rejected": -1.1015068292617798, + "step": 750 + }, + { + "epoch": 0.18234165067178504, + "grad_norm": 15.98215793865638, + "learning_rate": 4.897549265378004e-07, + "logits/chosen": 0.5963197946548462, + "logits/rejected": 0.6680821180343628, + "logps/chosen": -442.1761779785156, + "logps/rejected": -464.924072265625, + "loss": 0.5878, + "rewards/accuracies": 0.625, + "rewards/chosen": -0.9059039950370789, + "rewards/margins": 0.32418501377105713, + "rewards/rejected": -1.2300891876220703, + "step": 760 + }, + { + "epoch": 0.1847408829174664, + "grad_norm": 19.34691799566141, + "learning_rate": 4.891532581443643e-07, + "logits/chosen": 0.480471134185791, + "logits/rejected": 0.4600948691368103, + "logps/chosen": -396.38311767578125, + "logps/rejected": -465.755126953125, + "loss": 0.568, + "rewards/accuracies": 0.75, + "rewards/chosen": -0.6615660786628723, + "rewards/margins": 0.7741894721984863, + "rewards/rejected": -1.4357554912567139, + "step": 770 + }, + { + "epoch": 0.1871401151631478, + "grad_norm": 21.646866570334893, + "learning_rate": 4.885348141000122e-07, + "logits/chosen": 0.38621702790260315, + "logits/rejected": 0.41974037885665894, + "logps/chosen": -340.9786071777344, + "logps/rejected": -393.6761779785156, + "loss": 0.5754, + "rewards/accuracies": 0.75, + "rewards/chosen": -0.7244313955307007, + "rewards/margins": 0.48300299048423767, + "rewards/rejected": -1.2074342966079712, + "step": 780 + }, + { + "epoch": 0.18953934740882916, + "grad_norm": 18.58507989863696, + "learning_rate": 4.878996377861367e-07, + "logits/chosen": 0.39947646856307983, + "logits/rejected": 0.4246433675289154, + "logps/chosen": -320.23309326171875, + "logps/rejected": -370.582275390625, + "loss": 0.5467, + "rewards/accuracies": 0.699999988079071, + "rewards/chosen": -0.79985511302948, + "rewards/margins": 0.4771038591861725, + "rewards/rejected": -1.2769590616226196, + "step": 790 + }, + { + "epoch": 0.19193857965451055, + "grad_norm": 20.35280429241367, + "learning_rate": 4.872477737578327e-07, + "logits/chosen": 0.592739999294281, + "logits/rejected": 0.6939637660980225, + "logps/chosen": -408.095947265625, + "logps/rejected": -502.061279296875, + "loss": 0.5441, + "rewards/accuracies": 0.800000011920929, + "rewards/chosen": -0.9894254803657532, + "rewards/margins": 1.0690934658050537, + "rewards/rejected": -2.058518886566162, + "step": 800 + }, + { + "epoch": 0.19433781190019195, + "grad_norm": 26.00633546734353, + "learning_rate": 4.865792677407718e-07, + "logits/chosen": 0.35590919852256775, + "logits/rejected": 0.37258046865463257, + "logps/chosen": -347.53271484375, + "logps/rejected": -367.4189453125, + "loss": 0.5755, + "rewards/accuracies": 0.625, + "rewards/chosen": -0.8420926928520203, + "rewards/margins": 0.4168141484260559, + "rewards/rejected": -1.2589069604873657, + "step": 810 + }, + { + "epoch": 0.1967370441458733, + "grad_norm": 23.898332577859907, + "learning_rate": 4.858941666279955e-07, + "logits/chosen": 0.3112553358078003, + "logits/rejected": 0.35031309723854065, + "logps/chosen": -382.4200439453125, + "logps/rejected": -385.7547912597656, + "loss": 0.6063, + "rewards/accuracies": 0.6000000238418579, + "rewards/chosen": -0.877010703086853, + "rewards/margins": 0.20091959834098816, + "rewards/rejected": -1.0779304504394531, + "step": 820 + }, + { + "epoch": 0.1991362763915547, + "grad_norm": 19.075991074028558, + "learning_rate": 4.851925184766247e-07, + "logits/chosen": 0.3584078848361969, + "logits/rejected": 0.4070889949798584, + "logps/chosen": -362.82464599609375, + "logps/rejected": -392.70263671875, + "loss": 0.5809, + "rewards/accuracies": 0.7749999761581421, + "rewards/chosen": -0.8209225535392761, + "rewards/margins": 0.5464334487915039, + "rewards/rejected": -1.3673560619354248, + "step": 830 + }, + { + "epoch": 0.20153550863723607, + "grad_norm": 21.07807830955708, + "learning_rate": 4.844743725044897e-07, + "logits/chosen": 0.45863986015319824, + "logits/rejected": 0.4786900579929352, + "logps/chosen": -357.07562255859375, + "logps/rejected": -361.09991455078125, + "loss": 0.5894, + "rewards/accuracies": 0.6499999761581421, + "rewards/chosen": -0.8064752817153931, + "rewards/margins": 0.33143192529678345, + "rewards/rejected": -1.1379071474075317, + "step": 840 + }, + { + "epoch": 0.20393474088291746, + "grad_norm": 18.051907454620828, + "learning_rate": 4.837397790866774e-07, + "logits/chosen": 0.3541174530982971, + "logits/rejected": 0.38320186734199524, + "logps/chosen": -377.2417297363281, + "logps/rejected": -423.63092041015625, + "loss": 0.5831, + "rewards/accuracies": 0.75, + "rewards/chosen": -0.5865625143051147, + "rewards/margins": 0.7768771648406982, + "rewards/rejected": -1.3634397983551025, + "step": 850 + }, + { + "epoch": 0.20633397312859886, + "grad_norm": 22.343500765452667, + "learning_rate": 4.829887897519974e-07, + "logits/chosen": 0.38612139225006104, + "logits/rejected": 0.4179624617099762, + "logps/chosen": -313.2272033691406, + "logps/rejected": -380.7272033691406, + "loss": 0.5894, + "rewards/accuracies": 0.75, + "rewards/chosen": -0.6393442749977112, + "rewards/margins": 0.4778464436531067, + "rewards/rejected": -1.1171907186508179, + "step": 860 + }, + { + "epoch": 0.20873320537428022, + "grad_norm": 17.19888281834333, + "learning_rate": 4.82221457179368e-07, + "logits/chosen": 0.38400429487228394, + "logits/rejected": 0.4022274911403656, + "logps/chosen": -363.2855224609375, + "logps/rejected": -448.18951416015625, + "loss": 0.5606, + "rewards/accuracies": 0.800000011920929, + "rewards/chosen": -0.6849557757377625, + "rewards/margins": 1.044877529144287, + "rewards/rejected": -1.7298332452774048, + "step": 870 + }, + { + "epoch": 0.21113243761996162, + "grad_norm": 20.20842626295358, + "learning_rate": 4.814378351941206e-07, + "logits/chosen": 0.42639732360839844, + "logits/rejected": 0.41247129440307617, + "logps/chosen": -354.7998046875, + "logps/rejected": -369.0785827636719, + "loss": 0.5869, + "rewards/accuracies": 0.7749999761581421, + "rewards/chosen": -0.6992335915565491, + "rewards/margins": 0.3163672387599945, + "rewards/rejected": -1.0156008005142212, + "step": 880 + }, + { + "epoch": 0.21353166986564298, + "grad_norm": 19.600972978229347, + "learning_rate": 4.806379787642241e-07, + "logits/chosen": 0.5787872672080994, + "logits/rejected": 0.5056115388870239, + "logps/chosen": -356.230712890625, + "logps/rejected": -428.4977111816406, + "loss": 0.6034, + "rewards/accuracies": 0.7250000238418579, + "rewards/chosen": -0.8336772918701172, + "rewards/margins": 0.7834388017654419, + "rewards/rejected": -1.6171162128448486, + "step": 890 + }, + { + "epoch": 0.21593090211132437, + "grad_norm": 18.631627455895288, + "learning_rate": 4.798219439964293e-07, + "logits/chosen": 0.5892560482025146, + "logits/rejected": 0.6063620448112488, + "logps/chosen": -355.5292053222656, + "logps/rejected": -388.40380859375, + "loss": 0.5522, + "rewards/accuracies": 0.675000011920929, + "rewards/chosen": -0.7971435785293579, + "rewards/margins": 0.25393742322921753, + "rewards/rejected": -1.0510809421539307, + "step": 900 + }, + { + "epoch": 0.21833013435700577, + "grad_norm": 17.022161111644575, + "learning_rate": 4.78989788132333e-07, + "logits/chosen": 0.46557942032814026, + "logits/rejected": 0.48981013894081116, + "logps/chosen": -307.93316650390625, + "logps/rejected": -381.57257080078125, + "loss": 0.534, + "rewards/accuracies": 0.7250000238418579, + "rewards/chosen": -0.6179824471473694, + "rewards/margins": 0.7625503540039062, + "rewards/rejected": -1.3805327415466309, + "step": 910 + }, + { + "epoch": 0.22072936660268713, + "grad_norm": 20.84884348212129, + "learning_rate": 4.781415695443631e-07, + "logits/chosen": 0.4536976218223572, + "logits/rejected": 0.4783639907836914, + "logps/chosen": -462.7845764160156, + "logps/rejected": -509.302490234375, + "loss": 0.5778, + "rewards/accuracies": 0.574999988079071, + "rewards/chosen": -1.3845326900482178, + "rewards/margins": 0.4757843613624573, + "rewards/rejected": -1.8603169918060303, + "step": 920 + }, + { + "epoch": 0.22312859884836853, + "grad_norm": 17.49618245187623, + "learning_rate": 4.772773477316836e-07, + "logits/chosen": 0.432155042886734, + "logits/rejected": 0.4641633927822113, + "logps/chosen": -375.1294250488281, + "logps/rejected": -427.8644104003906, + "loss": 0.5509, + "rewards/accuracies": 0.75, + "rewards/chosen": -0.8244571685791016, + "rewards/margins": 0.5350643992424011, + "rewards/rejected": -1.3595216274261475, + "step": 930 + }, + { + "epoch": 0.2255278310940499, + "grad_norm": 22.293080856613333, + "learning_rate": 4.7639718331602117e-07, + "logits/chosen": 0.5172699093818665, + "logits/rejected": 0.5119847655296326, + "logps/chosen": -402.71343994140625, + "logps/rejected": -483.53826904296875, + "loss": 0.5483, + "rewards/accuracies": 0.8500000238418579, + "rewards/chosen": -0.9808033108711243, + "rewards/margins": 0.9733982086181641, + "rewards/rejected": -1.9542014598846436, + "step": 940 + }, + { + "epoch": 0.22792706333973128, + "grad_norm": 25.62636571756774, + "learning_rate": 4.7550113803741275e-07, + "logits/chosen": 0.5554766058921814, + "logits/rejected": 0.49316105246543884, + "logps/chosen": -416.37054443359375, + "logps/rejected": -398.11151123046875, + "loss": 0.5679, + "rewards/accuracies": 0.625, + "rewards/chosen": -0.9982360601425171, + "rewards/margins": 0.581201434135437, + "rewards/rejected": -1.5794376134872437, + "step": 950 + }, + { + "epoch": 0.23032629558541268, + "grad_norm": 23.381466639480767, + "learning_rate": 4.7458927474987454e-07, + "logits/chosen": 0.4861147403717041, + "logits/rejected": 0.47837033867836, + "logps/chosen": -430.23321533203125, + "logps/rejected": -403.3463439941406, + "loss": 0.5616, + "rewards/accuracies": 0.5249999761581421, + "rewards/chosen": -0.9009541273117065, + "rewards/margins": 0.359774112701416, + "rewards/rejected": -1.260728120803833, + "step": 960 + }, + { + "epoch": 0.23272552783109404, + "grad_norm": 18.23110087924843, + "learning_rate": 4.7366165741699347e-07, + "logits/chosen": 0.5253990888595581, + "logits/rejected": 0.4784235954284668, + "logps/chosen": -447.5752868652344, + "logps/rejected": -473.0809631347656, + "loss": 0.5516, + "rewards/accuracies": 0.625, + "rewards/chosen": -0.9549976587295532, + "rewards/margins": 0.6409324407577515, + "rewards/rejected": -1.5959299802780151, + "step": 970 + }, + { + "epoch": 0.23512476007677544, + "grad_norm": 19.550183415249666, + "learning_rate": 4.727183511074401e-07, + "logits/chosen": 0.45140281319618225, + "logits/rejected": 0.5012277960777283, + "logps/chosen": -384.2790832519531, + "logps/rejected": -401.28509521484375, + "loss": 0.5483, + "rewards/accuracies": 0.6499999761581421, + "rewards/chosen": -0.8066235780715942, + "rewards/margins": 0.3241332173347473, + "rewards/rejected": -1.1307567358016968, + "step": 980 + }, + { + "epoch": 0.2375239923224568, + "grad_norm": 20.922330115257477, + "learning_rate": 4.717594219904043e-07, + "logits/chosen": 0.43720340728759766, + "logits/rejected": 0.5218846797943115, + "logps/chosen": -401.09423828125, + "logps/rejected": -409.0234375, + "loss": 0.5863, + "rewards/accuracies": 0.675000011920929, + "rewards/chosen": -1.0844993591308594, + "rewards/margins": 0.5277708768844604, + "rewards/rejected": -1.6122701168060303, + "step": 990 + }, + { + "epoch": 0.2399232245681382, + "grad_norm": 20.924843509053616, + "learning_rate": 4.7078493733095393e-07, + "logits/chosen": 0.5131865739822388, + "logits/rejected": 0.5137609243392944, + "logps/chosen": -407.2503356933594, + "logps/rejected": -483.7713928222656, + "loss": 0.5376, + "rewards/accuracies": 0.699999988079071, + "rewards/chosen": -1.2562562227249146, + "rewards/margins": 0.7707854509353638, + "rewards/rejected": -2.0270419120788574, + "step": 1000 + }, + { + "epoch": 0.2423224568138196, + "grad_norm": 22.18855281521568, + "learning_rate": 4.6979496548531614e-07, + "logits/chosen": 0.6776459217071533, + "logits/rejected": 0.6498035788536072, + "logps/chosen": -403.9708557128906, + "logps/rejected": -523.07177734375, + "loss": 0.5635, + "rewards/accuracies": 0.675000011920929, + "rewards/chosen": -1.1393872499465942, + "rewards/margins": 0.7758584022521973, + "rewards/rejected": -1.9152456521987915, + "step": 1010 + }, + { + "epoch": 0.24472168905950095, + "grad_norm": 21.52071416914918, + "learning_rate": 4.6878957589608293e-07, + "logits/chosen": 0.5453814268112183, + "logits/rejected": 0.6066008806228638, + "logps/chosen": -425.591796875, + "logps/rejected": -567.7387084960938, + "loss": 0.5724, + "rewards/accuracies": 0.699999988079071, + "rewards/chosen": -1.3399444818496704, + "rewards/margins": 1.0568122863769531, + "rewards/rejected": -2.396756887435913, + "step": 1020 + }, + { + "epoch": 0.24712092130518235, + "grad_norm": 20.60341994179207, + "learning_rate": 4.6776883908733956e-07, + "logits/chosen": 0.6149223446846008, + "logits/rejected": 0.6715623736381531, + "logps/chosen": -438.75714111328125, + "logps/rejected": -463.2662048339844, + "loss": 0.5286, + "rewards/accuracies": 0.7749999761581421, + "rewards/chosen": -1.1835300922393799, + "rewards/margins": 0.9701372981071472, + "rewards/rejected": -2.153667449951172, + "step": 1030 + }, + { + "epoch": 0.2495201535508637, + "grad_norm": 28.539316703811078, + "learning_rate": 4.667328266597178e-07, + "logits/chosen": 0.5007244348526001, + "logits/rejected": 0.5865780711174011, + "logps/chosen": -385.4056701660156, + "logps/rejected": -439.2000427246094, + "loss": 0.531, + "rewards/accuracies": 0.699999988079071, + "rewards/chosen": -1.0280896425247192, + "rewards/margins": 0.6651249527931213, + "rewards/rejected": -1.6932144165039062, + "step": 1040 + }, + { + "epoch": 0.2519193857965451, + "grad_norm": 19.865896292426036, + "learning_rate": 4.6568161128537354e-07, + "logits/chosen": 0.40418606996536255, + "logits/rejected": 0.3820621371269226, + "logps/chosen": -416.3582458496094, + "logps/rejected": -402.7046203613281, + "loss": 0.5401, + "rewards/accuracies": 0.625, + "rewards/chosen": -1.3050010204315186, + "rewards/margins": 0.48114338517189026, + "rewards/rejected": -1.786144495010376, + "step": 1050 + }, + { + "epoch": 0.2543186180422265, + "grad_norm": 24.85634399900053, + "learning_rate": 4.6461526670288877e-07, + "logits/chosen": 0.5298096537590027, + "logits/rejected": 0.5856395959854126, + "logps/chosen": -398.751708984375, + "logps/rejected": -430.79644775390625, + "loss": 0.5911, + "rewards/accuracies": 0.675000011920929, + "rewards/chosen": -0.9989835619926453, + "rewards/margins": 0.6439894437789917, + "rewards/rejected": -1.6429731845855713, + "step": 1060 + }, + { + "epoch": 0.2567178502879079, + "grad_norm": 20.86937531865178, + "learning_rate": 4.635338677120994e-07, + "logits/chosen": 0.5040396451950073, + "logits/rejected": 0.5275683999061584, + "logps/chosen": -379.1663513183594, + "logps/rejected": -476.4314880371094, + "loss": 0.5262, + "rewards/accuracies": 0.75, + "rewards/chosen": -0.9181860089302063, + "rewards/margins": 0.8310756683349609, + "rewards/rejected": -1.749261498451233, + "step": 1070 + }, + { + "epoch": 0.2591170825335892, + "grad_norm": 23.1819970794382, + "learning_rate": 4.6243749016884835e-07, + "logits/chosen": 0.49801892042160034, + "logits/rejected": 0.5010608434677124, + "logps/chosen": -438.33074951171875, + "logps/rejected": -599.0955810546875, + "loss": 0.5475, + "rewards/accuracies": 0.8500000238418579, + "rewards/chosen": -1.3764212131500244, + "rewards/margins": 1.2255936861038208, + "rewards/rejected": -2.6020150184631348, + "step": 1080 + }, + { + "epoch": 0.2615163147792706, + "grad_norm": 30.911649534486866, + "learning_rate": 4.613262109796645e-07, + "logits/chosen": 0.43349236249923706, + "logits/rejected": 0.5261619687080383, + "logps/chosen": -420.1570739746094, + "logps/rejected": -584.062744140625, + "loss": 0.5436, + "rewards/accuracies": 0.7749999761581421, + "rewards/chosen": -1.3903928995132446, + "rewards/margins": 1.3632363080978394, + "rewards/rejected": -2.753629207611084, + "step": 1090 + }, + { + "epoch": 0.263915547024952, + "grad_norm": 24.211694093238176, + "learning_rate": 4.602001080963678e-07, + "logits/chosen": 0.21989259123802185, + "logits/rejected": 0.2885194420814514, + "logps/chosen": -407.5654296875, + "logps/rejected": -481.28448486328125, + "loss": 0.5503, + "rewards/accuracies": 0.675000011920929, + "rewards/chosen": -1.104230523109436, + "rewards/margins": 1.0199384689331055, + "rewards/rejected": -2.124168872833252, + "step": 1100 + }, + { + "epoch": 0.2663147792706334, + "grad_norm": 25.049981314231243, + "learning_rate": 4.590592605106017e-07, + "logits/chosen": 0.3270975947380066, + "logits/rejected": 0.33125847578048706, + "logps/chosen": -427.4678649902344, + "logps/rejected": -485.11590576171875, + "loss": 0.5704, + "rewards/accuracies": 0.8500000238418579, + "rewards/chosen": -1.1138917207717896, + "rewards/margins": 0.8963919878005981, + "rewards/rejected": -2.0102837085723877, + "step": 1110 + }, + { + "epoch": 0.2687140115163148, + "grad_norm": 21.939690926454343, + "learning_rate": 4.5790374824829165e-07, + "logits/chosen": 0.4945638179779053, + "logits/rejected": 0.4895549416542053, + "logps/chosen": -299.2240295410156, + "logps/rejected": -380.4999084472656, + "loss": 0.5841, + "rewards/accuracies": 0.800000011920929, + "rewards/chosen": -0.8892961740493774, + "rewards/margins": 0.7147836089134216, + "rewards/rejected": -1.6040798425674438, + "step": 1120 + }, + { + "epoch": 0.27111324376199614, + "grad_norm": 23.8591260962465, + "learning_rate": 4.5673365236403216e-07, + "logits/chosen": 0.4953368604183197, + "logits/rejected": 0.5128699541091919, + "logps/chosen": -334.7842712402344, + "logps/rejected": -490.0814514160156, + "loss": 0.5363, + "rewards/accuracies": 0.699999988079071, + "rewards/chosen": -1.113507628440857, + "rewards/margins": 1.3171967267990112, + "rewards/rejected": -2.430704355239868, + "step": 1130 + }, + { + "epoch": 0.27351247600767753, + "grad_norm": 20.28733876953354, + "learning_rate": 4.5554905493540075e-07, + "logits/chosen": 0.46798864006996155, + "logits/rejected": 0.5210005044937134, + "logps/chosen": -332.076904296875, + "logps/rejected": -446.70513916015625, + "loss": 0.5292, + "rewards/accuracies": 0.824999988079071, + "rewards/chosen": -0.9099678993225098, + "rewards/margins": 1.131079912185669, + "rewards/rejected": -2.041048049926758, + "step": 1140 + }, + { + "epoch": 0.2759117082533589, + "grad_norm": 24.87749514003154, + "learning_rate": 4.5435003905720074e-07, + "logits/chosen": 0.3501337170600891, + "logits/rejected": 0.44954681396484375, + "logps/chosen": -417.104736328125, + "logps/rejected": -438.91363525390625, + "loss": 0.5595, + "rewards/accuracies": 0.7250000238418579, + "rewards/chosen": -1.1025865077972412, + "rewards/margins": 0.5667014122009277, + "rewards/rejected": -1.669287919998169, + "step": 1150 + }, + { + "epoch": 0.2783109404990403, + "grad_norm": 27.415480953957644, + "learning_rate": 4.531366888356324e-07, + "logits/chosen": 0.47420987486839294, + "logits/rejected": 0.5211541652679443, + "logps/chosen": -319.5274353027344, + "logps/rejected": -489.9903869628906, + "loss": 0.5349, + "rewards/accuracies": 0.7749999761581421, + "rewards/chosen": -1.1082285642623901, + "rewards/margins": 1.3356571197509766, + "rewards/rejected": -2.443885564804077, + "step": 1160 + }, + { + "epoch": 0.2807101727447217, + "grad_norm": 29.146901804579972, + "learning_rate": 4.519090893823931e-07, + "logits/chosen": 0.4965229034423828, + "logits/rejected": 0.5144367814064026, + "logps/chosen": -405.8136291503906, + "logps/rejected": -467.6885681152344, + "loss": 0.5439, + "rewards/accuracies": 0.675000011920929, + "rewards/chosen": -1.3548388481140137, + "rewards/margins": 0.7498202919960022, + "rewards/rejected": -2.10465931892395, + "step": 1170 + }, + { + "epoch": 0.28310940499040305, + "grad_norm": 22.234128106063636, + "learning_rate": 4.5066732680870734e-07, + "logits/chosen": 0.4783782362937927, + "logits/rejected": 0.5079411864280701, + "logps/chosen": -382.99554443359375, + "logps/rejected": -473.92413330078125, + "loss": 0.5039, + "rewards/accuracies": 0.8500000238418579, + "rewards/chosen": -1.059499979019165, + "rewards/margins": 1.3445496559143066, + "rewards/rejected": -2.4040493965148926, + "step": 1180 + }, + { + "epoch": 0.28550863723608444, + "grad_norm": 26.81987709091395, + "learning_rate": 4.494114882192862e-07, + "logits/chosen": 0.3504638969898224, + "logits/rejected": 0.40155625343322754, + "logps/chosen": -391.93609619140625, + "logps/rejected": -485.8604431152344, + "loss": 0.519, + "rewards/accuracies": 0.8999999761581421, + "rewards/chosen": -1.090312123298645, + "rewards/margins": 1.276283860206604, + "rewards/rejected": -2.366596221923828, + "step": 1190 + }, + { + "epoch": 0.28790786948176583, + "grad_norm": 28.479329727395974, + "learning_rate": 4.4814166170621735e-07, + "logits/chosen": 0.4352048337459564, + "logits/rejected": 0.439187228679657, + "logps/chosen": -401.4508361816406, + "logps/rejected": -479.16046142578125, + "loss": 0.5459, + "rewards/accuracies": 0.7749999761581421, + "rewards/chosen": -1.2591068744659424, + "rewards/margins": 1.0630078315734863, + "rewards/rejected": -2.3221144676208496, + "step": 1200 + }, + { + "epoch": 0.2903071017274472, + "grad_norm": 24.36280340279443, + "learning_rate": 4.468579363427858e-07, + "logits/chosen": 0.3423922657966614, + "logits/rejected": 0.3217991888523102, + "logps/chosen": -408.8718566894531, + "logps/rejected": -476.89349365234375, + "loss": 0.5321, + "rewards/accuracies": 0.6499999761581421, + "rewards/chosen": -1.2676352262496948, + "rewards/margins": 1.0237175226211548, + "rewards/rejected": -2.2913527488708496, + "step": 1210 + }, + { + "epoch": 0.2927063339731286, + "grad_norm": 23.976543809274407, + "learning_rate": 4.4556040217722555e-07, + "logits/chosen": 0.4693296551704407, + "logits/rejected": 0.4371146261692047, + "logps/chosen": -361.5362548828125, + "logps/rejected": -458.41119384765625, + "loss": 0.5267, + "rewards/accuracies": 0.75, + "rewards/chosen": -0.9556007385253906, + "rewards/margins": 0.8090912103652954, + "rewards/rejected": -1.764691948890686, + "step": 1220 + }, + { + "epoch": 0.29510556621880996, + "grad_norm": 26.124294099974634, + "learning_rate": 4.442491502264033e-07, + "logits/chosen": 0.33999913930892944, + "logits/rejected": 0.3430488705635071, + "logps/chosen": -358.56488037109375, + "logps/rejected": -402.3939514160156, + "loss": 0.5327, + "rewards/accuracies": 0.625, + "rewards/chosen": -1.095595121383667, + "rewards/margins": 0.5801312327384949, + "rewards/rejected": -1.675726294517517, + "step": 1230 + }, + { + "epoch": 0.29750479846449135, + "grad_norm": 20.33193428165945, + "learning_rate": 4.429242724694338e-07, + "logits/chosen": 0.4010500907897949, + "logits/rejected": 0.40270981192588806, + "logps/chosen": -356.3854064941406, + "logps/rejected": -456.36859130859375, + "loss": 0.5449, + "rewards/accuracies": 0.699999988079071, + "rewards/chosen": -0.9084610939025879, + "rewards/margins": 0.8726455569267273, + "rewards/rejected": -1.7811065912246704, + "step": 1240 + }, + { + "epoch": 0.29990403071017274, + "grad_norm": 20.611238253632102, + "learning_rate": 4.4158586184122817e-07, + "logits/chosen": 0.42909201979637146, + "logits/rejected": 0.4563824236392975, + "logps/chosen": -440.75994873046875, + "logps/rejected": -490.83575439453125, + "loss": 0.5272, + "rewards/accuracies": 0.75, + "rewards/chosen": -1.2483009099960327, + "rewards/margins": 0.8814948201179504, + "rewards/rejected": -2.129795789718628, + "step": 1250 + }, + { + "epoch": 0.30230326295585414, + "grad_norm": 20.981763353845743, + "learning_rate": 4.4023401222597443e-07, + "logits/chosen": 0.4369926452636719, + "logits/rejected": 0.4608924984931946, + "logps/chosen": -429.01239013671875, + "logps/rejected": -467.69207763671875, + "loss": 0.5476, + "rewards/accuracies": 0.7749999761581421, + "rewards/chosen": -1.2999002933502197, + "rewards/margins": 0.5992451906204224, + "rewards/rejected": -1.899145483970642, + "step": 1260 + }, + { + "epoch": 0.30470249520153553, + "grad_norm": 27.79991213917318, + "learning_rate": 4.3886881845055235e-07, + "logits/chosen": 0.35042619705200195, + "logits/rejected": 0.3954378664493561, + "logps/chosen": -357.94622802734375, + "logps/rejected": -463.93359375, + "loss": 0.5244, + "rewards/accuracies": 0.800000011920929, + "rewards/chosen": -0.8437250852584839, + "rewards/margins": 1.2400966882705688, + "rewards/rejected": -2.0838217735290527, + "step": 1270 + }, + { + "epoch": 0.30710172744721687, + "grad_norm": 23.81723686863117, + "learning_rate": 4.374903762778814e-07, + "logits/chosen": 0.3485955595970154, + "logits/rejected": 0.41709309816360474, + "logps/chosen": -413.65496826171875, + "logps/rejected": -482.4605407714844, + "loss": 0.5143, + "rewards/accuracies": 0.699999988079071, + "rewards/chosen": -1.308464765548706, + "rewards/margins": 0.9804670214653015, + "rewards/rejected": -2.2889316082000732, + "step": 1280 + }, + { + "epoch": 0.30950095969289826, + "grad_norm": 26.981005902457373, + "learning_rate": 4.3609878240020356e-07, + "logits/chosen": 0.21841764450073242, + "logits/rejected": 0.3175281882286072, + "logps/chosen": -473.5174865722656, + "logps/rejected": -490.5992126464844, + "loss": 0.5435, + "rewards/accuracies": 0.675000011920929, + "rewards/chosen": -1.4544774293899536, + "rewards/margins": 0.8454269170761108, + "rewards/rejected": -2.2999043464660645, + "step": 1290 + }, + { + "epoch": 0.31190019193857965, + "grad_norm": 25.852778743297964, + "learning_rate": 4.346941344323005e-07, + "logits/chosen": 0.4441685676574707, + "logits/rejected": 0.5196029543876648, + "logps/chosen": -418.32501220703125, + "logps/rejected": -412.47015380859375, + "loss": 0.5668, + "rewards/accuracies": 0.7250000238418579, + "rewards/chosen": -1.4650938510894775, + "rewards/margins": 0.5241371989250183, + "rewards/rejected": -1.9892311096191406, + "step": 1300 + }, + { + "epoch": 0.31429942418426104, + "grad_norm": 24.34023435468119, + "learning_rate": 4.332765309046467e-07, + "logits/chosen": 0.5028194785118103, + "logits/rejected": 0.5491822361946106, + "logps/chosen": -414.00372314453125, + "logps/rejected": -485.53900146484375, + "loss": 0.5399, + "rewards/accuracies": 0.7749999761581421, + "rewards/chosen": -1.2709230184555054, + "rewards/margins": 1.1440144777297974, + "rewards/rejected": -2.4149374961853027, + "step": 1310 + }, + { + "epoch": 0.31669865642994244, + "grad_norm": 34.608951645650336, + "learning_rate": 4.3184607125649754e-07, + "logits/chosen": 0.5384331941604614, + "logits/rejected": 0.5625046491622925, + "logps/chosen": -426.2798767089844, + "logps/rejected": -554.5643310546875, + "loss": 0.548, + "rewards/accuracies": 0.800000011920929, + "rewards/chosen": -1.1276309490203857, + "rewards/margins": 1.2486127614974976, + "rewards/rejected": -2.3762435913085938, + "step": 1320 + }, + { + "epoch": 0.3190978886756238, + "grad_norm": 21.349543094305762, + "learning_rate": 4.304028558289141e-07, + "logits/chosen": 0.4867774546146393, + "logits/rejected": 0.5194205045700073, + "logps/chosen": -423.35272216796875, + "logps/rejected": -517.1748046875, + "loss": 0.5056, + "rewards/accuracies": 0.675000011920929, + "rewards/chosen": -1.104402780532837, + "rewards/margins": 1.1897287368774414, + "rewards/rejected": -2.2941317558288574, + "step": 1330 + }, + { + "epoch": 0.32149712092130517, + "grad_norm": 22.932371419845182, + "learning_rate": 4.28946985857725e-07, + "logits/chosen": 0.6204196214675903, + "logits/rejected": 0.6645633578300476, + "logps/chosen": -423.0869140625, + "logps/rejected": -528.7664184570312, + "loss": 0.5155, + "rewards/accuracies": 0.7749999761581421, + "rewards/chosen": -1.224035382270813, + "rewards/margins": 1.2077516317367554, + "rewards/rejected": -2.4317872524261475, + "step": 1340 + }, + { + "epoch": 0.32389635316698656, + "grad_norm": 25.20618572072171, + "learning_rate": 4.2747856346642445e-07, + "logits/chosen": 0.3865453600883484, + "logits/rejected": 0.4525006413459778, + "logps/chosen": -353.2498474121094, + "logps/rejected": -409.4380798339844, + "loss": 0.4923, + "rewards/accuracies": 0.800000011920929, + "rewards/chosen": -0.9723296165466309, + "rewards/margins": 0.8267601728439331, + "rewards/rejected": -1.799089789390564, + "step": 1350 + }, + { + "epoch": 0.32629558541266795, + "grad_norm": 38.18407998689627, + "learning_rate": 4.2599769165900933e-07, + "logits/chosen": 0.4781351685523987, + "logits/rejected": 0.5323950052261353, + "logps/chosen": -413.47967529296875, + "logps/rejected": -474.1600036621094, + "loss": 0.5526, + "rewards/accuracies": 0.7250000238418579, + "rewards/chosen": -1.4541221857070923, + "rewards/margins": 0.8590675592422485, + "rewards/rejected": -2.313189744949341, + "step": 1360 + }, + { + "epoch": 0.32869481765834935, + "grad_norm": 24.09623950016077, + "learning_rate": 4.245044743127535e-07, + "logits/chosen": 0.5714893341064453, + "logits/rejected": 0.6005902290344238, + "logps/chosen": -408.2439270019531, + "logps/rejected": -502.76617431640625, + "loss": 0.5289, + "rewards/accuracies": 0.7250000238418579, + "rewards/chosen": -1.2594388723373413, + "rewards/margins": 1.0005667209625244, + "rewards/rejected": -2.260005474090576, + "step": 1370 + }, + { + "epoch": 0.3310940499040307, + "grad_norm": 28.6113600733398, + "learning_rate": 4.229990161709214e-07, + "logits/chosen": 0.6574729681015015, + "logits/rejected": 0.748490571975708, + "logps/chosen": -373.8766784667969, + "logps/rejected": -522.918701171875, + "loss": 0.5396, + "rewards/accuracies": 0.800000011920929, + "rewards/chosen": -1.1603968143463135, + "rewards/margins": 1.344420075416565, + "rewards/rejected": -2.504816770553589, + "step": 1380 + }, + { + "epoch": 0.3334932821497121, + "grad_norm": 22.790330638281535, + "learning_rate": 4.214814228354204e-07, + "logits/chosen": 0.5872513651847839, + "logits/rejected": 0.6414788961410522, + "logps/chosen": -417.19696044921875, + "logps/rejected": -574.2335815429688, + "loss": 0.5233, + "rewards/accuracies": 0.8999999761581421, + "rewards/chosen": -1.2055175304412842, + "rewards/margins": 1.7793638706207275, + "rewards/rejected": -2.9848814010620117, + "step": 1390 + }, + { + "epoch": 0.33589251439539347, + "grad_norm": 26.978273454954337, + "learning_rate": 4.1995180075939375e-07, + "logits/chosen": 0.5491480827331543, + "logits/rejected": 0.6279400587081909, + "logps/chosen": -430.430908203125, + "logps/rejected": -508.8794860839844, + "loss": 0.496, + "rewards/accuracies": 0.699999988079071, + "rewards/chosen": -1.3094874620437622, + "rewards/margins": 1.0253905057907104, + "rewards/rejected": -2.3348779678344727, + "step": 1400 + }, + { + "epoch": 0.33829174664107486, + "grad_norm": 24.715640254579753, + "learning_rate": 4.1841025723975297e-07, + "logits/chosen": 0.4640998840332031, + "logits/rejected": 0.5079622268676758, + "logps/chosen": -419.351806640625, + "logps/rejected": -523.3448486328125, + "loss": 0.5036, + "rewards/accuracies": 0.8999999761581421, + "rewards/chosen": -1.0731154680252075, + "rewards/margins": 1.3017503023147583, + "rewards/rejected": -2.374865770339966, + "step": 1410 + }, + { + "epoch": 0.34069097888675626, + "grad_norm": 26.4759501785462, + "learning_rate": 4.168569004096516e-07, + "logits/chosen": 0.4973793029785156, + "logits/rejected": 0.4787571430206299, + "logps/chosen": -374.91082763671875, + "logps/rejected": -518.9568481445312, + "loss": 0.52, + "rewards/accuracies": 0.8500000238418579, + "rewards/chosen": -1.1492761373519897, + "rewards/margins": 1.374403953552246, + "rewards/rejected": -2.5236802101135254, + "step": 1420 + }, + { + "epoch": 0.3430902111324376, + "grad_norm": 24.37371579733604, + "learning_rate": 4.152918392308997e-07, + "logits/chosen": 0.6322020888328552, + "logits/rejected": 0.676570475101471, + "logps/chosen": -413.09967041015625, + "logps/rejected": -491.43280029296875, + "loss": 0.5219, + "rewards/accuracies": 0.7250000238418579, + "rewards/chosen": -1.3825113773345947, + "rewards/margins": 0.9496244192123413, + "rewards/rejected": -2.3321359157562256, + "step": 1430 + }, + { + "epoch": 0.345489443378119, + "grad_norm": 33.069428752057235, + "learning_rate": 4.137151834863213e-07, + "logits/chosen": 0.6233787536621094, + "logits/rejected": 0.6681519150733948, + "logps/chosen": -405.11572265625, + "logps/rejected": -605.8905029296875, + "loss": 0.5355, + "rewards/accuracies": 0.699999988079071, + "rewards/chosen": -1.4363985061645508, + "rewards/margins": 1.7567567825317383, + "rewards/rejected": -3.193154811859131, + "step": 1440 + }, + { + "epoch": 0.3478886756238004, + "grad_norm": 28.260532330049596, + "learning_rate": 4.121270437720526e-07, + "logits/chosen": 0.5473383665084839, + "logits/rejected": 0.5997258424758911, + "logps/chosen": -350.90234375, + "logps/rejected": -436.4149475097656, + "loss": 0.5328, + "rewards/accuracies": 0.75, + "rewards/chosen": -1.217137098312378, + "rewards/margins": 0.5057491064071655, + "rewards/rejected": -1.722886323928833, + "step": 1450 + }, + { + "epoch": 0.3502879078694818, + "grad_norm": 22.284111196687974, + "learning_rate": 4.105275314897852e-07, + "logits/chosen": 0.6474301815032959, + "logits/rejected": 0.660786509513855, + "logps/chosen": -404.047607421875, + "logps/rejected": -641.9457397460938, + "loss": 0.5045, + "rewards/accuracies": 0.75, + "rewards/chosen": -1.5259559154510498, + "rewards/margins": 2.1374330520629883, + "rewards/rejected": -3.66338849067688, + "step": 1460 + }, + { + "epoch": 0.35268714011516317, + "grad_norm": 25.147753437467735, + "learning_rate": 4.089167588389508e-07, + "logits/chosen": 0.622474193572998, + "logits/rejected": 0.6864532828330994, + "logps/chosen": -511.6786193847656, + "logps/rejected": -565.4790649414062, + "loss": 0.5338, + "rewards/accuracies": 0.75, + "rewards/chosen": -1.4496666193008423, + "rewards/margins": 1.0919597148895264, + "rewards/rejected": -2.541626214981079, + "step": 1470 + }, + { + "epoch": 0.3550863723608445, + "grad_norm": 30.072916780925627, + "learning_rate": 4.072948388088515e-07, + "logits/chosen": 0.6727067232131958, + "logits/rejected": 0.6959071159362793, + "logps/chosen": -438.89453125, + "logps/rejected": -549.9530029296875, + "loss": 0.5541, + "rewards/accuracies": 0.675000011920929, + "rewards/chosen": -1.4524840116500854, + "rewards/margins": 1.1177465915679932, + "rewards/rejected": -2.570230722427368, + "step": 1480 + }, + { + "epoch": 0.3574856046065259, + "grad_norm": 28.079457804957293, + "learning_rate": 4.056618851707334e-07, + "logits/chosen": 0.49242812395095825, + "logits/rejected": 0.5447872877120972, + "logps/chosen": -427.9210510253906, + "logps/rejected": -566.0662841796875, + "loss": 0.4829, + "rewards/accuracies": 0.8999999761581421, + "rewards/chosen": -1.2132065296173096, + "rewards/margins": 1.4744932651519775, + "rewards/rejected": -2.687699794769287, + "step": 1490 + }, + { + "epoch": 0.3598848368522073, + "grad_norm": 26.037654165610427, + "learning_rate": 4.0401801246980675e-07, + "logits/chosen": 0.531968355178833, + "logits/rejected": 0.5523951053619385, + "logps/chosen": -384.9483337402344, + "logps/rejected": -471.51300048828125, + "loss": 0.5461, + "rewards/accuracies": 0.75, + "rewards/chosen": -1.4794952869415283, + "rewards/margins": 1.0920584201812744, + "rewards/rejected": -2.5715537071228027, + "step": 1500 + }, + { + "epoch": 0.3622840690978887, + "grad_norm": 26.5263500776438, + "learning_rate": 4.0236333601721043e-07, + "logits/chosen": 0.33142679929733276, + "logits/rejected": 0.3150361478328705, + "logps/chosen": -445.0096130371094, + "logps/rejected": -515.8156127929688, + "loss": 0.5579, + "rewards/accuracies": 0.7749999761581421, + "rewards/chosen": -1.3551602363586426, + "rewards/margins": 0.6439944505691528, + "rewards/rejected": -1.9991546869277954, + "step": 1510 + }, + { + "epoch": 0.3646833013435701, + "grad_norm": 23.152212301904747, + "learning_rate": 4.0069797188192364e-07, + "logits/chosen": 0.5838603973388672, + "logits/rejected": 0.6234208345413208, + "logps/chosen": -459.1851501464844, + "logps/rejected": -571.7786254882812, + "loss": 0.5142, + "rewards/accuracies": 0.699999988079071, + "rewards/chosen": -1.4499067068099976, + "rewards/margins": 1.524433970451355, + "rewards/rejected": -2.9743404388427734, + "step": 1520 + }, + { + "epoch": 0.3670825335892514, + "grad_norm": 25.927981383141496, + "learning_rate": 3.9902203688262417e-07, + "logits/chosen": 0.44739946722984314, + "logits/rejected": 0.41936665773391724, + "logps/chosen": -408.06536865234375, + "logps/rejected": -458.7378845214844, + "loss": 0.5131, + "rewards/accuracies": 0.8500000238418579, + "rewards/chosen": -1.203150749206543, + "rewards/margins": 0.6691651940345764, + "rewards/rejected": -1.8723161220550537, + "step": 1530 + }, + { + "epoch": 0.3694817658349328, + "grad_norm": 29.57680066351449, + "learning_rate": 3.9733564857949365e-07, + "logits/chosen": 0.4725145399570465, + "logits/rejected": 0.5294008255004883, + "logps/chosen": -469.10107421875, + "logps/rejected": -531.1735229492188, + "loss": 0.4879, + "rewards/accuracies": 0.7250000238418579, + "rewards/chosen": -1.3328118324279785, + "rewards/margins": 1.1033190488815308, + "rewards/rejected": -2.436131000518799, + "step": 1540 + }, + { + "epoch": 0.3718809980806142, + "grad_norm": 29.953731902113084, + "learning_rate": 3.9563892526597177e-07, + "logits/chosen": 0.5199874639511108, + "logits/rejected": 0.5354346036911011, + "logps/chosen": -342.24407958984375, + "logps/rejected": -429.68341064453125, + "loss": 0.5281, + "rewards/accuracies": 0.699999988079071, + "rewards/chosen": -0.9911921620368958, + "rewards/margins": 0.44350871443748474, + "rewards/rejected": -1.4347009658813477, + "step": 1550 + }, + { + "epoch": 0.3742802303262956, + "grad_norm": 22.682880403794634, + "learning_rate": 3.9393198596045795e-07, + "logits/chosen": 0.5494628548622131, + "logits/rejected": 0.5927091836929321, + "logps/chosen": -379.50164794921875, + "logps/rejected": -497.7547302246094, + "loss": 0.5288, + "rewards/accuracies": 0.75, + "rewards/chosen": -1.1960418224334717, + "rewards/margins": 1.1193139553070068, + "rewards/rejected": -2.3153557777404785, + "step": 1560 + }, + { + "epoch": 0.376679462571977, + "grad_norm": 22.940020552459323, + "learning_rate": 3.922149503979628e-07, + "logits/chosen": 0.65094393491745, + "logits/rejected": 0.6770426630973816, + "logps/chosen": -493.34033203125, + "logps/rejected": -730.2711181640625, + "loss": 0.5097, + "rewards/accuracies": 0.824999988079071, + "rewards/chosen": -1.918515920639038, + "rewards/margins": 2.4283804893493652, + "rewards/rejected": -4.346896171569824, + "step": 1570 + }, + { + "epoch": 0.3790786948176583, + "grad_norm": 25.89668875929957, + "learning_rate": 3.904879390217095e-07, + "logits/chosen": 0.5813716650009155, + "logits/rejected": 0.5926984548568726, + "logps/chosen": -397.17742919921875, + "logps/rejected": -445.9786682128906, + "loss": 0.5204, + "rewards/accuracies": 0.824999988079071, + "rewards/chosen": -1.2316924333572388, + "rewards/margins": 0.7258312702178955, + "rewards/rejected": -1.9575237035751343, + "step": 1580 + }, + { + "epoch": 0.3814779270633397, + "grad_norm": 28.005322379907067, + "learning_rate": 3.8875107297468463e-07, + "logits/chosen": 0.47907742857933044, + "logits/rejected": 0.5216917395591736, + "logps/chosen": -399.86456298828125, + "logps/rejected": -612.7337036132812, + "loss": 0.5327, + "rewards/accuracies": 0.8999999761581421, + "rewards/chosen": -1.352043628692627, + "rewards/margins": 1.677664041519165, + "rewards/rejected": -3.029707431793213, + "step": 1590 + }, + { + "epoch": 0.3838771593090211, + "grad_norm": 27.73704631088212, + "learning_rate": 3.87004474091141e-07, + "logits/chosen": 0.6081745624542236, + "logits/rejected": 0.583394467830658, + "logps/chosen": -376.72491455078125, + "logps/rejected": -477.5032653808594, + "loss": 0.535, + "rewards/accuracies": 0.800000011920929, + "rewards/chosen": -1.322222113609314, + "rewards/margins": 0.8721188306808472, + "rewards/rejected": -2.194340944290161, + "step": 1600 + }, + { + "epoch": 0.3862763915547025, + "grad_norm": 26.265554246609113, + "learning_rate": 3.8524826488805114e-07, + "logits/chosen": 0.6204389333724976, + "logits/rejected": 0.6838095188140869, + "logps/chosen": -449.0771484375, + "logps/rejected": -474.48614501953125, + "loss": 0.5822, + "rewards/accuracies": 0.699999988079071, + "rewards/chosen": -1.383966326713562, + "rewards/margins": 0.7637938261032104, + "rewards/rejected": -2.1477601528167725, + "step": 1610 + }, + { + "epoch": 0.3886756238003839, + "grad_norm": 29.348434987188984, + "learning_rate": 3.834825685565133e-07, + "logits/chosen": 0.6158981323242188, + "logits/rejected": 0.6594232320785522, + "logps/chosen": -355.5142517089844, + "logps/rejected": -373.458740234375, + "loss": 0.4978, + "rewards/accuracies": 0.7250000238418579, + "rewards/chosen": -0.9645657539367676, + "rewards/margins": 0.5557066202163696, + "rewards/rejected": -1.5202723741531372, + "step": 1620 + }, + { + "epoch": 0.39107485604606523, + "grad_norm": 27.893225157055635, + "learning_rate": 3.8170750895311007e-07, + "logits/chosen": 0.5417597889900208, + "logits/rejected": 0.5491499900817871, + "logps/chosen": -412.00408935546875, + "logps/rejected": -480.2135314941406, + "loss": 0.4927, + "rewards/accuracies": 0.7749999761581421, + "rewards/chosen": -1.0488494634628296, + "rewards/margins": 0.9427961111068726, + "rewards/rejected": -1.9916454553604126, + "step": 1630 + }, + { + "epoch": 0.3934740882917466, + "grad_norm": 27.523271320708126, + "learning_rate": 3.7992321059122045e-07, + "logits/chosen": 0.6551875472068787, + "logits/rejected": 0.6813570857048035, + "logps/chosen": -417.86602783203125, + "logps/rejected": -490.6324768066406, + "loss": 0.5156, + "rewards/accuracies": 0.625, + "rewards/chosen": -1.4586915969848633, + "rewards/margins": 0.9774697422981262, + "rewards/rejected": -2.436161518096924, + "step": 1640 + }, + { + "epoch": 0.395873320537428, + "grad_norm": 27.996097883290435, + "learning_rate": 3.7812979863228576e-07, + "logits/chosen": 0.5485438108444214, + "logits/rejected": 0.5184262990951538, + "logps/chosen": -395.86700439453125, + "logps/rejected": -491.7772521972656, + "loss": 0.483, + "rewards/accuracies": 0.7250000238418579, + "rewards/chosen": -1.6062934398651123, + "rewards/margins": 0.8541345596313477, + "rewards/rejected": -2.46042799949646, + "step": 1650 + }, + { + "epoch": 0.3982725527831094, + "grad_norm": 29.83078441321164, + "learning_rate": 3.763273988770296e-07, + "logits/chosen": 0.604858934879303, + "logits/rejected": 0.6166650056838989, + "logps/chosen": -440.02288818359375, + "logps/rejected": -557.9898681640625, + "loss": 0.4932, + "rewards/accuracies": 0.824999988079071, + "rewards/chosen": -1.7407538890838623, + "rewards/margins": 1.2068676948547363, + "rewards/rejected": -2.9476218223571777, + "step": 1660 + }, + { + "epoch": 0.4006717850287908, + "grad_norm": 31.550622486136575, + "learning_rate": 3.7451613775663405e-07, + "logits/chosen": 0.5251548886299133, + "logits/rejected": 0.5625226497650146, + "logps/chosen": -412.3245544433594, + "logps/rejected": -622.3465576171875, + "loss": 0.5339, + "rewards/accuracies": 0.800000011920929, + "rewards/chosen": -1.4821490049362183, + "rewards/margins": 2.1304643154144287, + "rewards/rejected": -3.6126132011413574, + "step": 1670 + }, + { + "epoch": 0.40307101727447214, + "grad_norm": 36.283592176451926, + "learning_rate": 3.726961423238706e-07, + "logits/chosen": 0.49148011207580566, + "logits/rejected": 0.5424791574478149, + "logps/chosen": -365.7886962890625, + "logps/rejected": -534.2448120117188, + "loss": 0.5214, + "rewards/accuracies": 0.75, + "rewards/chosen": -1.1752984523773193, + "rewards/margins": 1.5028009414672852, + "rewards/rejected": -2.6780993938446045, + "step": 1680 + }, + { + "epoch": 0.40547024952015354, + "grad_norm": 28.465184644606882, + "learning_rate": 3.708675402441882e-07, + "logits/chosen": 0.5210837721824646, + "logits/rejected": 0.559233546257019, + "logps/chosen": -434.03173828125, + "logps/rejected": -449.0094299316406, + "loss": 0.5446, + "rewards/accuracies": 0.6000000238418579, + "rewards/chosen": -1.2292797565460205, + "rewards/margins": 0.6202489733695984, + "rewards/rejected": -1.8495289087295532, + "step": 1690 + }, + { + "epoch": 0.40786948176583493, + "grad_norm": 26.71337589674436, + "learning_rate": 3.6903045978675775e-07, + "logits/chosen": 0.5467537045478821, + "logits/rejected": 0.5603493452072144, + "logps/chosen": -381.9140930175781, + "logps/rejected": -523.8021240234375, + "loss": 0.5094, + "rewards/accuracies": 0.800000011920929, + "rewards/chosen": -1.217504858970642, + "rewards/margins": 1.6663074493408203, + "rewards/rejected": -2.883812427520752, + "step": 1700 + }, + { + "epoch": 0.4102687140115163, + "grad_norm": 21.803578934970087, + "learning_rate": 3.6718502981547474e-07, + "logits/chosen": 0.5511429905891418, + "logits/rejected": 0.5506604909896851, + "logps/chosen": -393.09063720703125, + "logps/rejected": -503.272705078125, + "loss": 0.5255, + "rewards/accuracies": 0.699999988079071, + "rewards/chosen": -1.1632798910140991, + "rewards/margins": 0.7249369025230408, + "rewards/rejected": -1.8882166147232056, + "step": 1710 + }, + { + "epoch": 0.4126679462571977, + "grad_norm": 23.93105714760726, + "learning_rate": 3.6533137977991986e-07, + "logits/chosen": 0.5443588495254517, + "logits/rejected": 0.6058872938156128, + "logps/chosen": -424.76654052734375, + "logps/rejected": -503.58905029296875, + "loss": 0.5524, + "rewards/accuracies": 0.7250000238418579, + "rewards/chosen": -1.1997069120407104, + "rewards/margins": 0.5387487411499023, + "rewards/rejected": -1.7384557723999023, + "step": 1720 + }, + { + "epoch": 0.41506717850287905, + "grad_norm": 23.947804764647334, + "learning_rate": 3.6346963970627865e-07, + "logits/chosen": 0.5812084078788757, + "logits/rejected": 0.66224205493927, + "logps/chosen": -391.9268493652344, + "logps/rejected": -514.671142578125, + "loss": 0.5117, + "rewards/accuracies": 0.7749999761581421, + "rewards/chosen": -1.2656793594360352, + "rewards/margins": 1.1044799089431763, + "rewards/rejected": -2.37015962600708, + "step": 1730 + }, + { + "epoch": 0.41746641074856045, + "grad_norm": 41.384778071002756, + "learning_rate": 3.615999401882207e-07, + "logits/chosen": 0.7019360065460205, + "logits/rejected": 0.7781286239624023, + "logps/chosen": -434.8446350097656, + "logps/rejected": -576.0948486328125, + "loss": 0.4974, + "rewards/accuracies": 0.7749999761581421, + "rewards/chosen": -1.8481266498565674, + "rewards/margins": 1.3344603776931763, + "rewards/rejected": -3.182587146759033, + "step": 1740 + }, + { + "epoch": 0.41986564299424184, + "grad_norm": 26.427988861693198, + "learning_rate": 3.597224123777389e-07, + "logits/chosen": 0.48930302262306213, + "logits/rejected": 0.5008028745651245, + "logps/chosen": -444.37188720703125, + "logps/rejected": -591.009521484375, + "loss": 0.5246, + "rewards/accuracies": 0.7250000238418579, + "rewards/chosen": -1.6480128765106201, + "rewards/margins": 1.4180339574813843, + "rewards/rejected": -3.066046714782715, + "step": 1750 + }, + { + "epoch": 0.42226487523992323, + "grad_norm": 28.520099902935833, + "learning_rate": 3.5783718797595e-07, + "logits/chosen": 0.5077411532402039, + "logits/rejected": 0.5465759634971619, + "logps/chosen": -451.1187438964844, + "logps/rejected": -498.052001953125, + "loss": 0.492, + "rewards/accuracies": 0.75, + "rewards/chosen": -1.2567503452301025, + "rewards/margins": 0.9746766090393066, + "rewards/rejected": -2.231426954269409, + "step": 1760 + }, + { + "epoch": 0.4246641074856046, + "grad_norm": 30.554823939974213, + "learning_rate": 3.559443992238558e-07, + "logits/chosen": 0.5454415082931519, + "logits/rejected": 0.5317539572715759, + "logps/chosen": -392.7830810546875, + "logps/rejected": -589.8780517578125, + "loss": 0.5362, + "rewards/accuracies": 0.875, + "rewards/chosen": -1.2405519485473633, + "rewards/margins": 1.7801573276519775, + "rewards/rejected": -3.020709276199341, + "step": 1770 + }, + { + "epoch": 0.42706333973128596, + "grad_norm": 29.80571254516348, + "learning_rate": 3.540441788930673e-07, + "logits/chosen": 0.4539203643798828, + "logits/rejected": 0.4859291613101959, + "logps/chosen": -468.7974548339844, + "logps/rejected": -593.902587890625, + "loss": 0.4975, + "rewards/accuracies": 0.800000011920929, + "rewards/chosen": -1.4451267719268799, + "rewards/margins": 1.7116448879241943, + "rewards/rejected": -3.156771659851074, + "step": 1780 + }, + { + "epoch": 0.42946257197696736, + "grad_norm": 25.06647355845889, + "learning_rate": 3.5213666027649123e-07, + "logits/chosen": 0.5843818783760071, + "logits/rejected": 0.6344080567359924, + "logps/chosen": -448.7802734375, + "logps/rejected": -481.861572265625, + "loss": 0.522, + "rewards/accuracies": 0.6000000238418579, + "rewards/chosen": -1.483563780784607, + "rewards/margins": 0.7844194173812866, + "rewards/rejected": -2.2679829597473145, + "step": 1790 + }, + { + "epoch": 0.43186180422264875, + "grad_norm": 36.388819937688396, + "learning_rate": 3.5022197717898017e-07, + "logits/chosen": 0.5060101747512817, + "logits/rejected": 0.5356935858726501, + "logps/chosen": -380.41058349609375, + "logps/rejected": -483.8663024902344, + "loss": 0.4632, + "rewards/accuracies": 0.675000011920929, + "rewards/chosen": -1.3735358715057373, + "rewards/margins": 1.4581668376922607, + "rewards/rejected": -2.831702709197998, + "step": 1800 + }, + { + "epoch": 0.43426103646833014, + "grad_norm": 29.77097075710584, + "learning_rate": 3.4830026390794633e-07, + "logits/chosen": 0.49516528844833374, + "logits/rejected": 0.48991116881370544, + "logps/chosen": -502.56036376953125, + "logps/rejected": -589.65234375, + "loss": 0.4919, + "rewards/accuracies": 0.75, + "rewards/chosen": -1.8622710704803467, + "rewards/margins": 1.399813175201416, + "rewards/rejected": -3.2620842456817627, + "step": 1810 + }, + { + "epoch": 0.43666026871401153, + "grad_norm": 24.214142767465663, + "learning_rate": 3.4637165526394104e-07, + "logits/chosen": 0.5962218642234802, + "logits/rejected": 0.5794373750686646, + "logps/chosen": -385.8623962402344, + "logps/rejected": -487.9881896972656, + "loss": 0.5179, + "rewards/accuracies": 0.7250000238418579, + "rewards/chosen": -1.3321025371551514, + "rewards/margins": 0.9605873823165894, + "rewards/rejected": -2.292689800262451, + "step": 1820 + }, + { + "epoch": 0.43905950095969287, + "grad_norm": 23.80846287144373, + "learning_rate": 3.4443628653119814e-07, + "logits/chosen": 0.4128933548927307, + "logits/rejected": 0.46047353744506836, + "logps/chosen": -453.2511291503906, + "logps/rejected": -729.1717529296875, + "loss": 0.5491, + "rewards/accuracies": 0.8500000238418579, + "rewards/chosen": -1.5917387008666992, + "rewards/margins": 2.3557941913604736, + "rewards/rejected": -3.947533369064331, + "step": 1830 + }, + { + "epoch": 0.44145873320537427, + "grad_norm": 26.22197096361496, + "learning_rate": 3.424942934681453e-07, + "logits/chosen": 0.3974430561065674, + "logits/rejected": 0.4507156312465668, + "logps/chosen": -367.8086853027344, + "logps/rejected": -498.85888671875, + "loss": 0.498, + "rewards/accuracies": 0.875, + "rewards/chosen": -1.061302900314331, + "rewards/margins": 1.3938474655151367, + "rewards/rejected": -2.4551501274108887, + "step": 1840 + }, + { + "epoch": 0.44385796545105566, + "grad_norm": 32.08569403406024, + "learning_rate": 3.405458122978804e-07, + "logits/chosen": 0.45793819427490234, + "logits/rejected": 0.4204599857330322, + "logps/chosen": -417.1183166503906, + "logps/rejected": -460.21307373046875, + "loss": 0.4997, + "rewards/accuracies": 0.800000011920929, + "rewards/chosen": -1.1381946802139282, + "rewards/margins": 0.8046306371688843, + "rewards/rejected": -1.9428255558013916, + "step": 1850 + }, + { + "epoch": 0.44625719769673705, + "grad_norm": 37.980694296967876, + "learning_rate": 3.3859097969861633e-07, + "logits/chosen": 0.5375388860702515, + "logits/rejected": 0.5443367958068848, + "logps/chosen": -446.30975341796875, + "logps/rejected": -508.3763732910156, + "loss": 0.539, + "rewards/accuracies": 0.699999988079071, + "rewards/chosen": -1.3490548133850098, + "rewards/margins": 1.0032390356063843, + "rewards/rejected": -2.3522937297821045, + "step": 1860 + }, + { + "epoch": 0.44865642994241844, + "grad_norm": 26.666657891859266, + "learning_rate": 3.366299327940936e-07, + "logits/chosen": 0.3666940927505493, + "logits/rejected": 0.43050676584243774, + "logps/chosen": -452.58447265625, + "logps/rejected": -585.009521484375, + "loss": 0.5032, + "rewards/accuracies": 0.7749999761581421, + "rewards/chosen": -1.3998275995254517, + "rewards/margins": 1.149751901626587, + "rewards/rejected": -2.54957914352417, + "step": 1870 + }, + { + "epoch": 0.4510556621880998, + "grad_norm": 24.096006375464146, + "learning_rate": 3.3466280914396117e-07, + "logits/chosen": 0.4738024175167084, + "logits/rejected": 0.5199214816093445, + "logps/chosen": -412.96148681640625, + "logps/rejected": -558.6272583007812, + "loss": 0.4846, + "rewards/accuracies": 0.675000011920929, + "rewards/chosen": -1.5531818866729736, + "rewards/margins": 1.2759530544281006, + "rewards/rejected": -2.829134702682495, + "step": 1880 + }, + { + "epoch": 0.4534548944337812, + "grad_norm": 33.47202296615829, + "learning_rate": 3.326897467341281e-07, + "logits/chosen": 0.5434667468070984, + "logits/rejected": 0.5950125455856323, + "logps/chosen": -384.5709228515625, + "logps/rejected": -510.760009765625, + "loss": 0.5062, + "rewards/accuracies": 0.699999988079071, + "rewards/chosen": -1.5631451606750488, + "rewards/margins": 1.1664178371429443, + "rewards/rejected": -2.7295632362365723, + "step": 1890 + }, + { + "epoch": 0.45585412667946257, + "grad_norm": 43.06627660738899, + "learning_rate": 3.3071088396708335e-07, + "logits/chosen": 0.43661123514175415, + "logits/rejected": 0.4971885681152344, + "logps/chosen": -351.27423095703125, + "logps/rejected": -520.4568481445312, + "loss": 0.5242, + "rewards/accuracies": 0.824999988079071, + "rewards/chosen": -1.249340534210205, + "rewards/margins": 1.5152721405029297, + "rewards/rejected": -2.7646126747131348, + "step": 1900 + }, + { + "epoch": 0.45825335892514396, + "grad_norm": 26.713110356503616, + "learning_rate": 3.2872635965218824e-07, + "logits/chosen": 0.5615112781524658, + "logits/rejected": 0.5642154216766357, + "logps/chosen": -469.04559326171875, + "logps/rejected": -581.6990966796875, + "loss": 0.5433, + "rewards/accuracies": 0.75, + "rewards/chosen": -1.8216552734375, + "rewards/margins": 1.0309664011001587, + "rewards/rejected": -2.8526217937469482, + "step": 1910 + }, + { + "epoch": 0.46065259117082535, + "grad_norm": 22.71250250046501, + "learning_rate": 3.2673631299593905e-07, + "logits/chosen": 0.47569403052330017, + "logits/rejected": 0.46433067321777344, + "logps/chosen": -456.00799560546875, + "logps/rejected": -559.7556762695312, + "loss": 0.5073, + "rewards/accuracies": 0.875, + "rewards/chosen": -1.5954478979110718, + "rewards/margins": 1.2076561450958252, + "rewards/rejected": -2.8031039237976074, + "step": 1920 + }, + { + "epoch": 0.4630518234165067, + "grad_norm": 37.7068693885991, + "learning_rate": 3.247408835922024e-07, + "logits/chosen": 0.32526105642318726, + "logits/rejected": 0.3423188328742981, + "logps/chosen": -531.8027954101562, + "logps/rejected": -669.9105224609375, + "loss": 0.5092, + "rewards/accuracies": 0.7250000238418579, + "rewards/chosen": -2.001676082611084, + "rewards/margins": 1.3426449298858643, + "rewards/rejected": -3.3443210124969482, + "step": 1930 + }, + { + "epoch": 0.4654510556621881, + "grad_norm": 28.132187213619453, + "learning_rate": 3.2274021141242306e-07, + "logits/chosen": 0.47985514998435974, + "logits/rejected": 0.5144001245498657, + "logps/chosen": -452.16876220703125, + "logps/rejected": -552.7636108398438, + "loss": 0.5115, + "rewards/accuracies": 0.7250000238418579, + "rewards/chosen": -1.7105381488800049, + "rewards/margins": 0.9188302755355835, + "rewards/rejected": -2.629368305206299, + "step": 1940 + }, + { + "epoch": 0.4678502879078695, + "grad_norm": 38.88584921816368, + "learning_rate": 3.2073443679580613e-07, + "logits/chosen": 0.35465139150619507, + "logits/rejected": 0.4013524055480957, + "logps/chosen": -436.73828125, + "logps/rejected": -477.15777587890625, + "loss": 0.5122, + "rewards/accuracies": 0.675000011920929, + "rewards/chosen": -1.45534086227417, + "rewards/margins": 0.4723455011844635, + "rewards/rejected": -1.927686333656311, + "step": 1950 + }, + { + "epoch": 0.47024952015355087, + "grad_norm": 22.941320708531656, + "learning_rate": 3.1872370043947194e-07, + "logits/chosen": 0.4294286370277405, + "logits/rejected": 0.4825877547264099, + "logps/chosen": -408.41009521484375, + "logps/rejected": -561.3411254882812, + "loss": 0.4806, + "rewards/accuracies": 0.824999988079071, + "rewards/chosen": -1.2056307792663574, + "rewards/margins": 1.601295828819275, + "rewards/rejected": -2.8069262504577637, + "step": 1960 + }, + { + "epoch": 0.47264875239923226, + "grad_norm": 35.45487358226262, + "learning_rate": 3.167081433885874e-07, + "logits/chosen": 0.6182829737663269, + "logits/rejected": 0.6686528921127319, + "logps/chosen": -513.27783203125, + "logps/rejected": -652.1695556640625, + "loss": 0.4855, + "rewards/accuracies": 0.699999988079071, + "rewards/chosen": -1.7456719875335693, + "rewards/margins": 1.04084050655365, + "rewards/rejected": -2.786512613296509, + "step": 1970 + }, + { + "epoch": 0.4750479846449136, + "grad_norm": 31.60040000561529, + "learning_rate": 3.14687907026472e-07, + "logits/chosen": 0.4577367901802063, + "logits/rejected": 0.47269877791404724, + "logps/chosen": -381.79400634765625, + "logps/rejected": -502.0397033691406, + "loss": 0.4878, + "rewards/accuracies": 0.800000011920929, + "rewards/chosen": -1.3730051517486572, + "rewards/margins": 1.0319459438323975, + "rewards/rejected": -2.404951333999634, + "step": 1980 + }, + { + "epoch": 0.477447216890595, + "grad_norm": 30.53844409715948, + "learning_rate": 3.126631330646801e-07, + "logits/chosen": 0.3548828959465027, + "logits/rejected": 0.3153901696205139, + "logps/chosen": -489.1748962402344, + "logps/rejected": -574.3113403320312, + "loss": 0.538, + "rewards/accuracies": 0.699999988079071, + "rewards/chosen": -1.6738046407699585, + "rewards/margins": 0.8269612193107605, + "rewards/rejected": -2.5007660388946533, + "step": 1990 + }, + { + "epoch": 0.4798464491362764, + "grad_norm": 26.31960630753036, + "learning_rate": 3.1063396353306097e-07, + "logits/chosen": 0.5535754561424255, + "logits/rejected": 0.571679949760437, + "logps/chosen": -414.44207763671875, + "logps/rejected": -463.46588134765625, + "loss": 0.5055, + "rewards/accuracies": 0.699999988079071, + "rewards/chosen": -1.2445614337921143, + "rewards/margins": 0.9598299860954285, + "rewards/rejected": -2.2043912410736084, + "step": 2000 + }, + { + "epoch": 0.4798464491362764, + "eval_logits/chosen": 0.5330603122711182, + "eval_logits/rejected": 0.5456880927085876, + "eval_logps/chosen": -420.8943786621094, + "eval_logps/rejected": -539.791015625, + "eval_loss": 0.5041713118553162, + "eval_rewards/accuracies": 0.7875000238418579, + "eval_rewards/chosen": -1.4196977615356445, + "eval_rewards/margins": 1.1845324039459229, + "eval_rewards/rejected": -2.6042304039001465, + "eval_runtime": 98.5276, + "eval_samples_per_second": 45.277, + "eval_steps_per_second": 0.71, + "step": 2000 + }, + { + "epoch": 0.4822456813819578, + "grad_norm": 36.92749686465627, + "learning_rate": 3.0860054076979535e-07, + "logits/chosen": 0.5472335815429688, + "logits/rejected": 0.5719302296638489, + "logps/chosen": -478.77862548828125, + "logps/rejected": -559.6343383789062, + "loss": 0.4953, + "rewards/accuracies": 0.800000011920929, + "rewards/chosen": -1.7711069583892822, + "rewards/margins": 1.1662440299987793, + "rewards/rejected": -2.9373512268066406, + "step": 2010 + }, + { + "epoch": 0.4846449136276392, + "grad_norm": 29.365570252191475, + "learning_rate": 3.065630074114115e-07, + "logits/chosen": 0.5482034087181091, + "logits/rejected": 0.5884445309638977, + "logps/chosen": -440.0167541503906, + "logps/rejected": -539.4930419921875, + "loss": 0.5277, + "rewards/accuracies": 0.7749999761581421, + "rewards/chosen": -1.330872893333435, + "rewards/margins": 1.472027063369751, + "rewards/rejected": -2.8028998374938965, + "step": 2020 + }, + { + "epoch": 0.4870441458733205, + "grad_norm": 26.28817793829581, + "learning_rate": 3.0452150638277947e-07, + "logits/chosen": 0.4989794194698334, + "logits/rejected": 0.5507176518440247, + "logps/chosen": -386.51409912109375, + "logps/rejected": -468.39410400390625, + "loss": 0.4966, + "rewards/accuracies": 0.699999988079071, + "rewards/chosen": -1.416949987411499, + "rewards/margins": 0.7112621665000916, + "rewards/rejected": -2.1282119750976562, + "step": 2030 + }, + { + "epoch": 0.4894433781190019, + "grad_norm": 31.006956473511085, + "learning_rate": 3.024761808870856e-07, + "logits/chosen": 0.6589056253433228, + "logits/rejected": 0.6739957332611084, + "logps/chosen": -375.49212646484375, + "logps/rejected": -606.4843139648438, + "loss": 0.4704, + "rewards/accuracies": 0.8999999761581421, + "rewards/chosen": -1.1116955280303955, + "rewards/margins": 2.418699026107788, + "rewards/rejected": -3.5303947925567627, + "step": 2040 + }, + { + "epoch": 0.4918426103646833, + "grad_norm": 45.236179424056324, + "learning_rate": 3.004271743957875e-07, + "logits/chosen": 0.495623916387558, + "logits/rejected": 0.501970112323761, + "logps/chosen": -460.5111389160156, + "logps/rejected": -548.1988525390625, + "loss": 0.5101, + "rewards/accuracies": 0.75, + "rewards/chosen": -1.778485655784607, + "rewards/margins": 0.600300669670105, + "rewards/rejected": -2.378786087036133, + "step": 2050 + }, + { + "epoch": 0.4942418426103647, + "grad_norm": 32.660373747646204, + "learning_rate": 2.983746306385499e-07, + "logits/chosen": 0.5202032923698425, + "logits/rejected": 0.5697475671768188, + "logps/chosen": -439.6302185058594, + "logps/rejected": -633.529296875, + "loss": 0.5057, + "rewards/accuracies": 0.800000011920929, + "rewards/chosen": -1.7451140880584717, + "rewards/margins": 1.8138139247894287, + "rewards/rejected": -3.5589280128479004, + "step": 2060 + }, + { + "epoch": 0.4966410748560461, + "grad_norm": 51.57954812756242, + "learning_rate": 2.963186935931628e-07, + "logits/chosen": 0.49899429082870483, + "logits/rejected": 0.5181163549423218, + "logps/chosen": -429.69696044921875, + "logps/rejected": -546.2327270507812, + "loss": 0.4849, + "rewards/accuracies": 0.8500000238418579, + "rewards/chosen": -1.4596073627471924, + "rewards/margins": 1.2351174354553223, + "rewards/rejected": -2.6947247982025146, + "step": 2070 + }, + { + "epoch": 0.4990403071017274, + "grad_norm": 37.743557346368355, + "learning_rate": 2.9425950747544176e-07, + "logits/chosen": 0.4497598111629486, + "logits/rejected": 0.4462924003601074, + "logps/chosen": -512.8709106445312, + "logps/rejected": -644.2141723632812, + "loss": 0.4828, + "rewards/accuracies": 0.75, + "rewards/chosen": -1.907044768333435, + "rewards/margins": 1.6575613021850586, + "rewards/rejected": -3.5646064281463623, + "step": 2080 + }, + { + "epoch": 0.5014395393474088, + "grad_norm": 38.732238817533585, + "learning_rate": 2.921972167291119e-07, + "logits/chosen": 0.6549265384674072, + "logits/rejected": 0.6424221396446228, + "logps/chosen": -469.24627685546875, + "logps/rejected": -607.3594970703125, + "loss": 0.4949, + "rewards/accuracies": 0.800000011920929, + "rewards/chosen": -1.6534907817840576, + "rewards/margins": 1.179262399673462, + "rewards/rejected": -2.8327534198760986, + "step": 2090 + }, + { + "epoch": 0.5038387715930902, + "grad_norm": 32.977127485837634, + "learning_rate": 2.9013196601567567e-07, + "logits/chosen": 0.22545738518238068, + "logits/rejected": 0.31531310081481934, + "logps/chosen": -394.68255615234375, + "logps/rejected": -499.6689453125, + "loss": 0.5539, + "rewards/accuracies": 0.8500000238418579, + "rewards/chosen": -1.201900839805603, + "rewards/margins": 0.9213203191757202, + "rewards/rejected": -2.123220920562744, + "step": 2100 + }, + { + "epoch": 0.5062380038387716, + "grad_norm": 31.676779604073975, + "learning_rate": 2.8806390020426555e-07, + "logits/chosen": 0.5821847915649414, + "logits/rejected": 0.6351027488708496, + "logps/chosen": -432.86981201171875, + "logps/rejected": -522.4436645507812, + "loss": 0.4991, + "rewards/accuracies": 0.824999988079071, + "rewards/chosen": -1.3247367143630981, + "rewards/margins": 0.9480254054069519, + "rewards/rejected": -2.2727620601654053, + "step": 2110 + }, + { + "epoch": 0.508637236084453, + "grad_norm": 35.00486375316812, + "learning_rate": 2.8599316436148187e-07, + "logits/chosen": 0.4525614380836487, + "logits/rejected": 0.4556906819343567, + "logps/chosen": -410.01922607421875, + "logps/rejected": -492.8228454589844, + "loss": 0.5013, + "rewards/accuracies": 0.675000011920929, + "rewards/chosen": -1.4033153057098389, + "rewards/margins": 0.8626031875610352, + "rewards/rejected": -2.265918493270874, + "step": 2120 + }, + { + "epoch": 0.5110364683301344, + "grad_norm": 30.270031114348164, + "learning_rate": 2.8391990374121723e-07, + "logits/chosen": 0.6305533647537231, + "logits/rejected": 0.6111758351325989, + "logps/chosen": -405.37762451171875, + "logps/rejected": -599.3681640625, + "loss": 0.5044, + "rewards/accuracies": 0.800000011920929, + "rewards/chosen": -1.399217963218689, + "rewards/margins": 1.6433988809585571, + "rewards/rejected": -3.042617082595825, + "step": 2130 + }, + { + "epoch": 0.5134357005758158, + "grad_norm": 25.427591996310284, + "learning_rate": 2.818442637744669e-07, + "logits/chosen": 0.5907707810401917, + "logits/rejected": 0.6095005869865417, + "logps/chosen": -420.5110778808594, + "logps/rejected": -531.2003784179688, + "loss": 0.5094, + "rewards/accuracies": 0.75, + "rewards/chosen": -1.4939762353897095, + "rewards/margins": 1.0677590370178223, + "rewards/rejected": -2.561735153198242, + "step": 2140 + }, + { + "epoch": 0.5158349328214972, + "grad_norm": 29.10370817287732, + "learning_rate": 2.797663900591284e-07, + "logits/chosen": 0.5958229303359985, + "logits/rejected": 0.5724313855171204, + "logps/chosen": -438.81842041015625, + "logps/rejected": -509.04364013671875, + "loss": 0.4796, + "rewards/accuracies": 0.7749999761581421, + "rewards/chosen": -1.548699975013733, + "rewards/margins": 1.0004830360412598, + "rewards/rejected": -2.5491831302642822, + "step": 2150 + }, + { + "epoch": 0.5182341650671785, + "grad_norm": 29.350472811549906, + "learning_rate": 2.776864283497874e-07, + "logits/chosen": 0.5696316361427307, + "logits/rejected": 0.5993034243583679, + "logps/chosen": -406.0208435058594, + "logps/rejected": -559.4927978515625, + "loss": 0.5002, + "rewards/accuracies": 0.625, + "rewards/chosen": -1.5112321376800537, + "rewards/margins": 1.6407535076141357, + "rewards/rejected": -3.1519858837127686, + "step": 2160 + }, + { + "epoch": 0.5206333973128598, + "grad_norm": 30.396592351915984, + "learning_rate": 2.756045245474943e-07, + "logits/chosen": 0.4575433135032654, + "logits/rejected": 0.5362235307693481, + "logps/chosen": -446.55767822265625, + "logps/rejected": -541.7037963867188, + "loss": 0.4922, + "rewards/accuracies": 0.824999988079071, + "rewards/chosen": -1.6036920547485352, + "rewards/margins": 0.8041864633560181, + "rewards/rejected": -2.4078783988952637, + "step": 2170 + }, + { + "epoch": 0.5230326295585412, + "grad_norm": 34.376914467887886, + "learning_rate": 2.7352082468952977e-07, + "logits/chosen": 0.6164439916610718, + "logits/rejected": 0.5993391871452332, + "logps/chosen": -476.3313903808594, + "logps/rejected": -687.8834228515625, + "loss": 0.554, + "rewards/accuracies": 0.699999988079071, + "rewards/chosen": -2.068695545196533, + "rewards/margins": 1.9827091693878174, + "rewards/rejected": -4.0514044761657715, + "step": 2180 + }, + { + "epoch": 0.5254318618042226, + "grad_norm": 32.201249898617604, + "learning_rate": 2.7143547493916e-07, + "logits/chosen": 0.6297436952590942, + "logits/rejected": 0.6931129693984985, + "logps/chosen": -443.54193115234375, + "logps/rejected": -651.564453125, + "loss": 0.48, + "rewards/accuracies": 0.7749999761581421, + "rewards/chosen": -1.6411831378936768, + "rewards/margins": 2.111694812774658, + "rewards/rejected": -3.752877712249756, + "step": 2190 + }, + { + "epoch": 0.527831094049904, + "grad_norm": 33.481424863253956, + "learning_rate": 2.693486215753853e-07, + "logits/chosen": 0.581652820110321, + "logits/rejected": 0.6015012860298157, + "logps/chosen": -468.0342712402344, + "logps/rejected": -665.8737182617188, + "loss": 0.5107, + "rewards/accuracies": 0.699999988079071, + "rewards/chosen": -1.9511600732803345, + "rewards/margins": 2.2348785400390625, + "rewards/rejected": -4.186038017272949, + "step": 2200 + }, + { + "epoch": 0.5302303262955854, + "grad_norm": 30.668163376222036, + "learning_rate": 2.6726041098267805e-07, + "logits/chosen": 0.5464416742324829, + "logits/rejected": 0.5375434160232544, + "logps/chosen": -479.27197265625, + "logps/rejected": -527.6573486328125, + "loss": 0.5522, + "rewards/accuracies": 0.625, + "rewards/chosen": -1.6696078777313232, + "rewards/margins": 0.8726032972335815, + "rewards/rejected": -2.5422110557556152, + "step": 2210 + }, + { + "epoch": 0.5326295585412668, + "grad_norm": 50.75005189933063, + "learning_rate": 2.6517098964071507e-07, + "logits/chosen": 0.6589860320091248, + "logits/rejected": 0.6931478977203369, + "logps/chosen": -399.00360107421875, + "logps/rejected": -464.0816955566406, + "loss": 0.5398, + "rewards/accuracies": 0.75, + "rewards/chosen": -1.2213267087936401, + "rewards/margins": 0.4684023857116699, + "rewards/rejected": -1.6897293329238892, + "step": 2220 + }, + { + "epoch": 0.5350287907869482, + "grad_norm": 41.032076102256525, + "learning_rate": 2.630805041141023e-07, + "logits/chosen": 0.7128033638000488, + "logits/rejected": 0.7279654741287231, + "logps/chosen": -364.8663024902344, + "logps/rejected": -559.3302001953125, + "loss": 0.4992, + "rewards/accuracies": 0.699999988079071, + "rewards/chosen": -1.1767915487289429, + "rewards/margins": 1.811647653579712, + "rewards/rejected": -2.9884393215179443, + "step": 2230 + }, + { + "epoch": 0.5374280230326296, + "grad_norm": 53.34540663235414, + "learning_rate": 2.609891010420941e-07, + "logits/chosen": 0.649802565574646, + "logits/rejected": 0.7241895198822021, + "logps/chosen": -443.50811767578125, + "logps/rejected": -610.3703002929688, + "loss": 0.4712, + "rewards/accuracies": 0.800000011920929, + "rewards/chosen": -1.4965871572494507, + "rewards/margins": 1.7070690393447876, + "rewards/rejected": -3.203655958175659, + "step": 2240 + }, + { + "epoch": 0.539827255278311, + "grad_norm": 25.357687908984758, + "learning_rate": 2.5889692712830674e-07, + "logits/chosen": 0.630439043045044, + "logits/rejected": 0.6549093127250671, + "logps/chosen": -375.9720153808594, + "logps/rejected": -464.39239501953125, + "loss": 0.496, + "rewards/accuracies": 0.824999988079071, + "rewards/chosen": -1.233820915222168, + "rewards/margins": 0.9906282424926758, + "rewards/rejected": -2.2244491577148438, + "step": 2250 + }, + { + "epoch": 0.5422264875239923, + "grad_norm": 47.499335691257166, + "learning_rate": 2.5680412913042843e-07, + "logits/chosen": 0.5655387043952942, + "logits/rejected": 0.6218944191932678, + "logps/chosen": -422.29351806640625, + "logps/rejected": -549.7659301757812, + "loss": 0.5181, + "rewards/accuracies": 0.800000011920929, + "rewards/chosen": -1.5186660289764404, + "rewards/margins": 1.3507541418075562, + "rewards/rejected": -2.869419813156128, + "step": 2260 + }, + { + "epoch": 0.5446257197696737, + "grad_norm": 27.17413580646397, + "learning_rate": 2.5471085384992404e-07, + "logits/chosen": 0.6208091974258423, + "logits/rejected": 0.6463747024536133, + "logps/chosen": -433.322265625, + "logps/rejected": -678.259521484375, + "loss": 0.4805, + "rewards/accuracies": 0.800000011920929, + "rewards/chosen": -1.6338462829589844, + "rewards/margins": 2.446549654006958, + "rewards/rejected": -4.080395698547363, + "step": 2270 + }, + { + "epoch": 0.5470249520153551, + "grad_norm": 28.316091683395783, + "learning_rate": 2.526172481217381e-07, + "logits/chosen": 0.5385035276412964, + "logits/rejected": 0.5358555316925049, + "logps/chosen": -412.51226806640625, + "logps/rejected": -599.2774658203125, + "loss": 0.5164, + "rewards/accuracies": 0.7250000238418579, + "rewards/chosen": -1.8855476379394531, + "rewards/margins": 1.7057392597198486, + "rewards/rejected": -3.591287136077881, + "step": 2280 + }, + { + "epoch": 0.5494241842610365, + "grad_norm": 34.541036258138135, + "learning_rate": 2.5052345880399456e-07, + "logits/chosen": 0.518667459487915, + "logits/rejected": 0.5254421830177307, + "logps/chosen": -403.07269287109375, + "logps/rejected": -506.0286560058594, + "loss": 0.4737, + "rewards/accuracies": 0.699999988079071, + "rewards/chosen": -1.5274192094802856, + "rewards/margins": 1.0553488731384277, + "rewards/rejected": -2.582767963409424, + "step": 2290 + }, + { + "epoch": 0.5518234165067178, + "grad_norm": 28.277531402961085, + "learning_rate": 2.4842963276769555e-07, + "logits/chosen": 0.5363879203796387, + "logits/rejected": 0.5436598062515259, + "logps/chosen": -434.55340576171875, + "logps/rejected": -601.2906494140625, + "loss": 0.5008, + "rewards/accuracies": 0.75, + "rewards/chosen": -1.9992926120758057, + "rewards/margins": 1.2217543125152588, + "rewards/rejected": -3.2210469245910645, + "step": 2300 + }, + { + "epoch": 0.5542226487523992, + "grad_norm": 28.240328527545266, + "learning_rate": 2.463359168864189e-07, + "logits/chosen": 0.4857005178928375, + "logits/rejected": 0.5038618445396423, + "logps/chosen": -473.5176696777344, + "logps/rejected": -521.3526611328125, + "loss": 0.5321, + "rewards/accuracies": 0.7250000238418579, + "rewards/chosen": -1.535585880279541, + "rewards/margins": 0.8739916086196899, + "rewards/rejected": -2.4095776081085205, + "step": 2310 + }, + { + "epoch": 0.5566218809980806, + "grad_norm": 28.035767914492325, + "learning_rate": 2.4424245802601555e-07, + "logits/chosen": 0.46195000410079956, + "logits/rejected": 0.4942271113395691, + "logps/chosen": -370.6980895996094, + "logps/rejected": -514.0704345703125, + "loss": 0.4942, + "rewards/accuracies": 0.7250000238418579, + "rewards/chosen": -1.2649797201156616, + "rewards/margins": 0.9138463735580444, + "rewards/rejected": -2.178826093673706, + "step": 2320 + }, + { + "epoch": 0.559021113243762, + "grad_norm": 49.81514108152477, + "learning_rate": 2.421494030343072e-07, + "logits/chosen": 0.6403766870498657, + "logits/rejected": 0.6394528150558472, + "logps/chosen": -436.5330505371094, + "logps/rejected": -459.36480712890625, + "loss": 0.5775, + "rewards/accuracies": 0.824999988079071, + "rewards/chosen": -1.452787160873413, + "rewards/margins": 0.7825342416763306, + "rewards/rejected": -2.235321283340454, + "step": 2330 + }, + { + "epoch": 0.5614203454894434, + "grad_norm": 39.02908540341096, + "learning_rate": 2.400568987307861e-07, + "logits/chosen": 0.647792398929596, + "logits/rejected": 0.5832428932189941, + "logps/chosen": -378.17205810546875, + "logps/rejected": -390.7867431640625, + "loss": 0.5065, + "rewards/accuracies": 0.75, + "rewards/chosen": -1.234156608581543, + "rewards/margins": 0.34215831756591797, + "rewards/rejected": -1.5763150453567505, + "step": 2340 + }, + { + "epoch": 0.5638195777351248, + "grad_norm": 34.25301402942352, + "learning_rate": 2.379650918963156e-07, + "logits/chosen": 0.5757008194923401, + "logits/rejected": 0.6183763742446899, + "logps/chosen": -355.47906494140625, + "logps/rejected": -486.69232177734375, + "loss": 0.5056, + "rewards/accuracies": 0.800000011920929, + "rewards/chosen": -1.4054075479507446, + "rewards/margins": 1.2384564876556396, + "rewards/rejected": -2.643864154815674, + "step": 2350 + }, + { + "epoch": 0.5662188099808061, + "grad_norm": 33.95997970816329, + "learning_rate": 2.3587412926283438e-07, + "logits/chosen": 0.5819541811943054, + "logits/rejected": 0.6103062033653259, + "logps/chosen": -490.56640625, + "logps/rejected": -584.3535766601562, + "loss": 0.5231, + "rewards/accuracies": 0.625, + "rewards/chosen": -1.5636104345321655, + "rewards/margins": 1.433632254600525, + "rewards/rejected": -2.9972424507141113, + "step": 2360 + }, + { + "epoch": 0.5686180422264875, + "grad_norm": 33.319507197288296, + "learning_rate": 2.337841575030642e-07, + "logits/chosen": 0.48387107253074646, + "logits/rejected": 0.48154598474502563, + "logps/chosen": -442.59228515625, + "logps/rejected": -581.6680908203125, + "loss": 0.5024, + "rewards/accuracies": 0.824999988079071, + "rewards/chosen": -1.3801770210266113, + "rewards/margins": 1.2666184902191162, + "rewards/rejected": -2.6467957496643066, + "step": 2370 + }, + { + "epoch": 0.5710172744721689, + "grad_norm": 31.154558612258484, + "learning_rate": 2.316953232202206e-07, + "logits/chosen": 0.7043946385383606, + "logits/rejected": 0.7090233564376831, + "logps/chosen": -413.64569091796875, + "logps/rejected": -415.6429138183594, + "loss": 0.487, + "rewards/accuracies": 0.699999988079071, + "rewards/chosen": -1.4569604396820068, + "rewards/margins": 0.7636488676071167, + "rewards/rejected": -2.220609188079834, + "step": 2380 + }, + { + "epoch": 0.5734165067178503, + "grad_norm": 27.53888081945469, + "learning_rate": 2.2960777293772958e-07, + "logits/chosen": 0.7477514147758484, + "logits/rejected": 0.7476789951324463, + "logps/chosen": -377.96209716796875, + "logps/rejected": -513.9478759765625, + "loss": 0.4799, + "rewards/accuracies": 0.7749999761581421, + "rewards/chosen": -1.4383244514465332, + "rewards/margins": 1.5858564376831055, + "rewards/rejected": -3.0241806507110596, + "step": 2390 + }, + { + "epoch": 0.5758157389635317, + "grad_norm": 29.519016591031086, + "learning_rate": 2.2752165308894974e-07, + "logits/chosen": 0.6164499521255493, + "logits/rejected": 0.6799559593200684, + "logps/chosen": -361.77264404296875, + "logps/rejected": -481.71002197265625, + "loss": 0.479, + "rewards/accuracies": 0.699999988079071, + "rewards/chosen": -1.447037935256958, + "rewards/margins": 1.3069896697998047, + "rewards/rejected": -2.7540273666381836, + "step": 2400 + }, + { + "epoch": 0.5782149712092131, + "grad_norm": 28.74044368904408, + "learning_rate": 2.254371100069005e-07, + "logits/chosen": 0.551243782043457, + "logits/rejected": 0.6056830286979675, + "logps/chosen": -397.4892578125, + "logps/rejected": -533.1715087890625, + "loss": 0.4819, + "rewards/accuracies": 0.824999988079071, + "rewards/chosen": -1.3637256622314453, + "rewards/margins": 1.1748710870742798, + "rewards/rejected": -2.5385966300964355, + "step": 2410 + }, + { + "epoch": 0.5806142034548945, + "grad_norm": 39.23052270442962, + "learning_rate": 2.2335428991399725e-07, + "logits/chosen": 0.7540609240531921, + "logits/rejected": 0.725174605846405, + "logps/chosen": -453.46246337890625, + "logps/rejected": -699.2947998046875, + "loss": 0.5017, + "rewards/accuracies": 0.7250000238418579, + "rewards/chosen": -2.230548620223999, + "rewards/margins": 2.478804588317871, + "rewards/rejected": -4.709353446960449, + "step": 2420 + }, + { + "epoch": 0.5830134357005758, + "grad_norm": 35.02504460778454, + "learning_rate": 2.2127333891179458e-07, + "logits/chosen": 0.5849272012710571, + "logits/rejected": 0.5850101113319397, + "logps/chosen": -398.72564697265625, + "logps/rejected": -597.7645874023438, + "loss": 0.5377, + "rewards/accuracies": 0.800000011920929, + "rewards/chosen": -1.6642696857452393, + "rewards/margins": 1.7866287231445312, + "rewards/rejected": -3.4508986473083496, + "step": 2430 + }, + { + "epoch": 0.5854126679462572, + "grad_norm": 34.70931185312274, + "learning_rate": 2.1919440297073782e-07, + "logits/chosen": 0.6965154409408569, + "logits/rejected": 0.6873563528060913, + "logps/chosen": -419.25323486328125, + "logps/rejected": -576.9484252929688, + "loss": 0.5063, + "rewards/accuracies": 0.75, + "rewards/chosen": -1.7598737478256226, + "rewards/margins": 1.6468613147735596, + "rewards/rejected": -3.4067349433898926, + "step": 2440 + }, + { + "epoch": 0.5878119001919386, + "grad_norm": 28.419435291801697, + "learning_rate": 2.1711762791992368e-07, + "logits/chosen": 0.6497688293457031, + "logits/rejected": 0.6952397227287292, + "logps/chosen": -469.9932556152344, + "logps/rejected": -512.4960327148438, + "loss": 0.5487, + "rewards/accuracies": 0.75, + "rewards/chosen": -1.6104280948638916, + "rewards/margins": 0.7736623883247375, + "rewards/rejected": -2.3840904235839844, + "step": 2450 + }, + { + "epoch": 0.5902111324376199, + "grad_norm": 33.62103705328426, + "learning_rate": 2.1504315943687114e-07, + "logits/chosen": 0.7049362659454346, + "logits/rejected": 0.7318881750106812, + "logps/chosen": -400.5235900878906, + "logps/rejected": -579.788330078125, + "loss": 0.5231, + "rewards/accuracies": 0.824999988079071, + "rewards/chosen": -1.4686295986175537, + "rewards/margins": 1.3642832040786743, + "rewards/rejected": -2.8329129219055176, + "step": 2460 + }, + { + "epoch": 0.5926103646833013, + "grad_norm": 37.1550051643393, + "learning_rate": 2.1297114303730248e-07, + "logits/chosen": 0.6910517811775208, + "logits/rejected": 0.7440094947814941, + "logps/chosen": -433.5245666503906, + "logps/rejected": -632.4937133789062, + "loss": 0.5607, + "rewards/accuracies": 0.675000011920929, + "rewards/chosen": -1.9925916194915771, + "rewards/margins": 1.4477274417877197, + "rewards/rejected": -3.440319061279297, + "step": 2470 + }, + { + "epoch": 0.5950095969289827, + "grad_norm": 32.702979787660254, + "learning_rate": 2.1090172406493616e-07, + "logits/chosen": 0.5501781702041626, + "logits/rejected": 0.6104339361190796, + "logps/chosen": -356.7936096191406, + "logps/rejected": -473.5712890625, + "loss": 0.4741, + "rewards/accuracies": 0.8999999761581421, + "rewards/chosen": -1.0627294778823853, + "rewards/margins": 1.095290184020996, + "rewards/rejected": -2.158019542694092, + "step": 2480 + }, + { + "epoch": 0.5974088291746641, + "grad_norm": 37.57322885787863, + "learning_rate": 2.0883504768129146e-07, + "logits/chosen": 0.6268981695175171, + "logits/rejected": 0.6416260600090027, + "logps/chosen": -450.3939514160156, + "logps/rejected": -575.8052978515625, + "loss": 0.5084, + "rewards/accuracies": 0.7749999761581421, + "rewards/chosen": -1.5946685075759888, + "rewards/margins": 1.3188021183013916, + "rewards/rejected": -2.91347074508667, + "step": 2490 + }, + { + "epoch": 0.5998080614203455, + "grad_norm": 39.256486442369265, + "learning_rate": 2.0677125885550571e-07, + "logits/chosen": 0.6842622756958008, + "logits/rejected": 0.6938222646713257, + "logps/chosen": -393.2452697753906, + "logps/rejected": -461.9776916503906, + "loss": 0.5113, + "rewards/accuracies": 0.7250000238418579, + "rewards/chosen": -1.3430721759796143, + "rewards/margins": 1.0853911638259888, + "rewards/rejected": -2.4284629821777344, + "step": 2500 + }, + { + "epoch": 0.6022072936660269, + "grad_norm": 30.68214318996349, + "learning_rate": 2.0471050235416587e-07, + "logits/chosen": 0.675729513168335, + "logits/rejected": 0.7008381485939026, + "logps/chosen": -472.68341064453125, + "logps/rejected": -557.2034912109375, + "loss": 0.4783, + "rewards/accuracies": 0.7749999761581421, + "rewards/chosen": -1.8632663488388062, + "rewards/margins": 1.3563127517700195, + "rewards/rejected": -3.219578981399536, + "step": 2510 + }, + { + "epoch": 0.6046065259117083, + "grad_norm": 31.3311200225744, + "learning_rate": 2.026529227311532e-07, + "logits/chosen": 0.6629911661148071, + "logits/rejected": 0.7233511805534363, + "logps/chosen": -439.72784423828125, + "logps/rejected": -584.7631225585938, + "loss": 0.528, + "rewards/accuracies": 0.699999988079071, + "rewards/chosen": -2.0253746509552, + "rewards/margins": 1.3782660961151123, + "rewards/rejected": -3.4036407470703125, + "step": 2520 + }, + { + "epoch": 0.6070057581573897, + "grad_norm": 35.26584973890508, + "learning_rate": 2.005986643175036e-07, + "logits/chosen": 0.6147924661636353, + "logits/rejected": 0.6977729797363281, + "logps/chosen": -460.37890625, + "logps/rejected": -651.292724609375, + "loss": 0.4456, + "rewards/accuracies": 0.8500000238418579, + "rewards/chosen": -1.7438169717788696, + "rewards/margins": 2.0838496685028076, + "rewards/rejected": -3.827666759490967, + "step": 2530 + }, + { + "epoch": 0.6094049904030711, + "grad_norm": 34.233093320124546, + "learning_rate": 1.9854787121128328e-07, + "logits/chosen": 0.6681974530220032, + "logits/rejected": 0.6297106742858887, + "logps/chosen": -375.5139465332031, + "logps/rejected": -458.5330505371094, + "loss": 0.5165, + "rewards/accuracies": 0.7749999761581421, + "rewards/chosen": -1.3370566368103027, + "rewards/margins": 1.3033640384674072, + "rewards/rejected": -2.640420913696289, + "step": 2540 + }, + { + "epoch": 0.6118042226487524, + "grad_norm": 28.920920445134406, + "learning_rate": 1.9650068726748106e-07, + "logits/chosen": 0.7325125932693481, + "logits/rejected": 0.7009168863296509, + "logps/chosen": -475.2723083496094, + "logps/rejected": -622.3833618164062, + "loss": 0.5236, + "rewards/accuracies": 0.800000011920929, + "rewards/chosen": -1.810741662979126, + "rewards/margins": 1.578853726387024, + "rewards/rejected": -3.3895955085754395, + "step": 2550 + }, + { + "epoch": 0.6142034548944337, + "grad_norm": 30.861207689091124, + "learning_rate": 1.9445725608791718e-07, + "logits/chosen": 0.7458786964416504, + "logits/rejected": 0.7599055171012878, + "logps/chosen": -476.2272033691406, + "logps/rejected": -782.8841552734375, + "loss": 0.4996, + "rewards/accuracies": 0.7749999761581421, + "rewards/chosen": -1.934945821762085, + "rewards/margins": 3.0150437355041504, + "rewards/rejected": -4.9499897956848145, + "step": 2560 + }, + { + "epoch": 0.6166026871401151, + "grad_norm": 31.420286094143293, + "learning_rate": 1.924177210111705e-07, + "logits/chosen": 0.6218976378440857, + "logits/rejected": 0.6657293438911438, + "logps/chosen": -420.7046813964844, + "logps/rejected": -610.647705078125, + "loss": 0.4977, + "rewards/accuracies": 0.824999988079071, + "rewards/chosen": -1.6917369365692139, + "rewards/margins": 1.8296884298324585, + "rewards/rejected": -3.521425724029541, + "step": 2570 + }, + { + "epoch": 0.6190019193857965, + "grad_norm": 30.509222434763775, + "learning_rate": 1.9038222510252364e-07, + "logits/chosen": 0.6954709887504578, + "logits/rejected": 0.777031421661377, + "logps/chosen": -401.15545654296875, + "logps/rejected": -490.4656677246094, + "loss": 0.5043, + "rewards/accuracies": 0.7250000238418579, + "rewards/chosen": -1.3204840421676636, + "rewards/margins": 1.0271937847137451, + "rewards/rejected": -2.3476779460906982, + "step": 2580 + }, + { + "epoch": 0.6214011516314779, + "grad_norm": 40.99082438692143, + "learning_rate": 1.883509111439277e-07, + "logits/chosen": 0.6641609072685242, + "logits/rejected": 0.7061329483985901, + "logps/chosen": -401.11053466796875, + "logps/rejected": -666.2782592773438, + "loss": 0.522, + "rewards/accuracies": 0.824999988079071, + "rewards/chosen": -1.5550295114517212, + "rewards/margins": 1.9435818195343018, + "rewards/rejected": -3.4986109733581543, + "step": 2590 + }, + { + "epoch": 0.6238003838771593, + "grad_norm": 23.80753562651165, + "learning_rate": 1.8632392162398665e-07, + "logits/chosen": 0.7196059226989746, + "logits/rejected": 0.7276310920715332, + "logps/chosen": -464.3755798339844, + "logps/rejected": -605.0327758789062, + "loss": 0.4924, + "rewards/accuracies": 0.925000011920929, + "rewards/chosen": -1.446529507637024, + "rewards/margins": 1.4860612154006958, + "rewards/rejected": -2.932590961456299, + "step": 2600 + }, + { + "epoch": 0.6261996161228407, + "grad_norm": 35.50656541014287, + "learning_rate": 1.84301398727962e-07, + "logits/chosen": 0.7047610282897949, + "logits/rejected": 0.7627818584442139, + "logps/chosen": -335.9312744140625, + "logps/rejected": -600.0538330078125, + "loss": 0.5143, + "rewards/accuracies": 0.8500000238418579, + "rewards/chosen": -1.3366457223892212, + "rewards/margins": 2.3122611045837402, + "rewards/rejected": -3.648906707763672, + "step": 2610 + }, + { + "epoch": 0.6285988483685221, + "grad_norm": 31.93713658952523, + "learning_rate": 1.8228348432779966e-07, + "logits/chosen": 0.5606223940849304, + "logits/rejected": 0.6245654821395874, + "logps/chosen": -408.01715087890625, + "logps/rejected": -522.1866455078125, + "loss": 0.5116, + "rewards/accuracies": 0.7749999761581421, + "rewards/chosen": -1.572877049446106, + "rewards/margins": 1.2301509380340576, + "rewards/rejected": -2.803027629852295, + "step": 2620 + }, + { + "epoch": 0.6309980806142035, + "grad_norm": 27.914653903784515, + "learning_rate": 1.8027031997217773e-07, + "logits/chosen": 0.7705515623092651, + "logits/rejected": 0.7918896675109863, + "logps/chosen": -443.3219299316406, + "logps/rejected": -657.1170654296875, + "loss": 0.4764, + "rewards/accuracies": 0.75, + "rewards/chosen": -2.029221773147583, + "rewards/margins": 2.0203018188476562, + "rewards/rejected": -4.04952335357666, + "step": 2630 + }, + { + "epoch": 0.6333973128598849, + "grad_norm": 24.442785917599732, + "learning_rate": 1.7826204687657758e-07, + "logits/chosen": 0.7035326957702637, + "logits/rejected": 0.721504807472229, + "logps/chosen": -443.2125549316406, + "logps/rejected": -473.4432678222656, + "loss": 0.4983, + "rewards/accuracies": 0.75, + "rewards/chosen": -1.4505491256713867, + "rewards/margins": 0.755431056022644, + "rewards/rejected": -2.205980062484741, + "step": 2640 + }, + { + "epoch": 0.6357965451055663, + "grad_norm": 31.963860621592772, + "learning_rate": 1.762588059133781e-07, + "logits/chosen": 0.7350751161575317, + "logits/rejected": 0.7096099853515625, + "logps/chosen": -475.07305908203125, + "logps/rejected": -598.759521484375, + "loss": 0.4743, + "rewards/accuracies": 0.8500000238418579, + "rewards/chosen": -1.5520153045654297, + "rewards/margins": 1.6064621210098267, + "rewards/rejected": -3.158477306365967, + "step": 2650 + }, + { + "epoch": 0.6381957773512476, + "grad_norm": 31.899547328781036, + "learning_rate": 1.7426073760197406e-07, + "logits/chosen": 0.6485757827758789, + "logits/rejected": 0.7398275136947632, + "logps/chosen": -451.94580078125, + "logps/rejected": -686.9503173828125, + "loss": 0.5078, + "rewards/accuracies": 0.824999988079071, + "rewards/chosen": -1.814279556274414, + "rewards/margins": 2.0318989753723145, + "rewards/rejected": -3.8461780548095703, + "step": 2660 + }, + { + "epoch": 0.6405950095969289, + "grad_norm": 26.178708516639514, + "learning_rate": 1.7226798209891935e-07, + "logits/chosen": 0.6045451760292053, + "logits/rejected": 0.6039005517959595, + "logps/chosen": -454.6619567871094, + "logps/rejected": -534.8763427734375, + "loss": 0.4857, + "rewards/accuracies": 0.8500000238418579, + "rewards/chosen": -1.7492996454238892, + "rewards/margins": 1.4018948078155518, + "rewards/rejected": -3.1511945724487305, + "step": 2670 + }, + { + "epoch": 0.6429942418426103, + "grad_norm": 35.91366947948363, + "learning_rate": 1.7028067918809535e-07, + "logits/chosen": 0.5839170217514038, + "logits/rejected": 0.6014143228530884, + "logps/chosen": -384.09564208984375, + "logps/rejected": -648.3985595703125, + "loss": 0.4921, + "rewards/accuracies": 0.875, + "rewards/chosen": -1.4092109203338623, + "rewards/margins": 2.2909350395202637, + "rewards/rejected": -3.700145721435547, + "step": 2680 + }, + { + "epoch": 0.6453934740882917, + "grad_norm": 35.647393846898865, + "learning_rate": 1.6829896827090584e-07, + "logits/chosen": 0.6380244493484497, + "logits/rejected": 0.6699076294898987, + "logps/chosen": -416.11297607421875, + "logps/rejected": -456.918212890625, + "loss": 0.5431, + "rewards/accuracies": 0.6499999761581421, + "rewards/chosen": -1.3569297790527344, + "rewards/margins": 0.7398675084114075, + "rewards/rejected": -2.096797227859497, + "step": 2690 + }, + { + "epoch": 0.6477927063339731, + "grad_norm": 21.784137715119133, + "learning_rate": 1.6632298835649844e-07, + "logits/chosen": 0.6910983920097351, + "logits/rejected": 0.6865051984786987, + "logps/chosen": -429.507080078125, + "logps/rejected": -623.1798706054688, + "loss": 0.454, + "rewards/accuracies": 0.7749999761581421, + "rewards/chosen": -1.3549401760101318, + "rewards/margins": 1.7029485702514648, + "rewards/rejected": -3.0578887462615967, + "step": 2700 + }, + { + "epoch": 0.6501919385796545, + "grad_norm": 40.680020887507624, + "learning_rate": 1.6435287805201364e-07, + "logits/chosen": 0.6894032955169678, + "logits/rejected": 0.6674355268478394, + "logps/chosen": -420.45233154296875, + "logps/rejected": -537.4315185546875, + "loss": 0.5036, + "rewards/accuracies": 0.800000011920929, + "rewards/chosen": -1.3876147270202637, + "rewards/margins": 1.1851856708526611, + "rewards/rejected": -2.572800397872925, + "step": 2710 + }, + { + "epoch": 0.6525911708253359, + "grad_norm": 35.99817855160938, + "learning_rate": 1.6238877555286207e-07, + "logits/chosen": 0.6698207855224609, + "logits/rejected": 0.7016371488571167, + "logps/chosen": -433.239501953125, + "logps/rejected": -564.9044799804688, + "loss": 0.4667, + "rewards/accuracies": 0.8500000238418579, + "rewards/chosen": -1.443068504333496, + "rewards/margins": 1.2255724668502808, + "rewards/rejected": -2.6686408519744873, + "step": 2720 + }, + { + "epoch": 0.6549904030710173, + "grad_norm": 39.75965895420522, + "learning_rate": 1.60430818633031e-07, + "logits/chosen": 0.6557576656341553, + "logits/rejected": 0.6801853775978088, + "logps/chosen": -389.7122497558594, + "logps/rejected": -529.9295654296875, + "loss": 0.4612, + "rewards/accuracies": 0.7749999761581421, + "rewards/chosen": -1.1586706638336182, + "rewards/margins": 1.4817416667938232, + "rewards/rejected": -2.6404123306274414, + "step": 2730 + }, + { + "epoch": 0.6573896353166987, + "grad_norm": 27.544765289204754, + "learning_rate": 1.5847914463541939e-07, + "logits/chosen": 0.6880534291267395, + "logits/rejected": 0.6456581950187683, + "logps/chosen": -367.79229736328125, + "logps/rejected": -525.629638671875, + "loss": 0.4842, + "rewards/accuracies": 0.75, + "rewards/chosen": -1.4672930240631104, + "rewards/margins": 1.4001801013946533, + "rewards/rejected": -2.8674731254577637, + "step": 2740 + }, + { + "epoch": 0.6597888675623801, + "grad_norm": 24.452787144356392, + "learning_rate": 1.5653389046220427e-07, + "logits/chosen": 0.5893281698226929, + "logits/rejected": 0.6441556215286255, + "logps/chosen": -364.4541931152344, + "logps/rejected": -486.20703125, + "loss": 0.5127, + "rewards/accuracies": 0.7749999761581421, + "rewards/chosen": -1.0919770002365112, + "rewards/margins": 1.092616319656372, + "rewards/rejected": -2.184593439102173, + "step": 2750 + }, + { + "epoch": 0.6621880998080614, + "grad_norm": 44.42792750388695, + "learning_rate": 1.545951925652375e-07, + "logits/chosen": 0.7141789197921753, + "logits/rejected": 0.7477242946624756, + "logps/chosen": -495.19659423828125, + "logps/rejected": -557.9563598632812, + "loss": 0.4907, + "rewards/accuracies": 0.800000011920929, + "rewards/chosen": -1.5828977823257446, + "rewards/margins": 1.3042711019515991, + "rewards/rejected": -2.8871684074401855, + "step": 2760 + }, + { + "epoch": 0.6645873320537428, + "grad_norm": 35.45188058330296, + "learning_rate": 1.5266318693647423e-07, + "logits/chosen": 0.7555517554283142, + "logits/rejected": 0.8401724100112915, + "logps/chosen": -423.3294372558594, + "logps/rejected": -492.0943908691406, + "loss": 0.4833, + "rewards/accuracies": 0.7250000238418579, + "rewards/chosen": -1.2794373035430908, + "rewards/margins": 0.7932703495025635, + "rewards/rejected": -2.0727076530456543, + "step": 2770 + }, + { + "epoch": 0.6669865642994242, + "grad_norm": 31.904169316096304, + "learning_rate": 1.5073800909843353e-07, + "logits/chosen": 0.6724790930747986, + "logits/rejected": 0.7021657824516296, + "logps/chosen": -463.25274658203125, + "logps/rejected": -532.02294921875, + "loss": 0.4598, + "rewards/accuracies": 0.75, + "rewards/chosen": -1.7021219730377197, + "rewards/margins": 1.30453622341156, + "rewards/rejected": -3.0066585540771484, + "step": 2780 + }, + { + "epoch": 0.6693857965451055, + "grad_norm": 43.15929522651468, + "learning_rate": 1.488197940946922e-07, + "logits/chosen": 0.6783190369606018, + "logits/rejected": 0.7536433935165405, + "logps/chosen": -446.2560119628906, + "logps/rejected": -520.4361572265625, + "loss": 0.4781, + "rewards/accuracies": 0.824999988079071, + "rewards/chosen": -1.470943570137024, + "rewards/margins": 1.3352140188217163, + "rewards/rejected": -2.8061575889587402, + "step": 2790 + }, + { + "epoch": 0.6717850287907869, + "grad_norm": 55.17192263922261, + "learning_rate": 1.4690867648041167e-07, + "logits/chosen": 0.6258468627929688, + "logits/rejected": 0.6363823413848877, + "logps/chosen": -445.82501220703125, + "logps/rejected": -570.8683471679688, + "loss": 0.4973, + "rewards/accuracies": 0.8500000238418579, + "rewards/chosen": -1.644078016281128, + "rewards/margins": 1.5276083946228027, + "rewards/rejected": -3.1716864109039307, + "step": 2800 + }, + { + "epoch": 0.6741842610364683, + "grad_norm": 37.27131536327267, + "learning_rate": 1.4500479031289987e-07, + "logits/chosen": 0.44763240218162537, + "logits/rejected": 0.46375036239624023, + "logps/chosen": -417.30914306640625, + "logps/rejected": -528.802734375, + "loss": 0.5146, + "rewards/accuracies": 0.8500000238418579, + "rewards/chosen": -1.2662084102630615, + "rewards/margins": 1.2129180431365967, + "rewards/rejected": -2.479126214981079, + "step": 2810 + }, + { + "epoch": 0.6765834932821497, + "grad_norm": 34.36417791117091, + "learning_rate": 1.4310826914220747e-07, + "logits/chosen": 0.5888969898223877, + "logits/rejected": 0.6028741002082825, + "logps/chosen": -482.534912109375, + "logps/rejected": -555.63330078125, + "loss": 0.5267, + "rewards/accuracies": 0.7250000238418579, + "rewards/chosen": -1.6587140560150146, + "rewards/margins": 0.8636137843132019, + "rewards/rejected": -2.5223278999328613, + "step": 2820 + }, + { + "epoch": 0.6789827255278311, + "grad_norm": 29.135677726916754, + "learning_rate": 1.412192460017597e-07, + "logits/chosen": 0.7223433256149292, + "logits/rejected": 0.7450364232063293, + "logps/chosen": -455.407470703125, + "logps/rejected": -618.5993041992188, + "loss": 0.513, + "rewards/accuracies": 0.7749999761581421, + "rewards/chosen": -1.9469377994537354, + "rewards/margins": 1.590118646621704, + "rewards/rejected": -3.5370564460754395, + "step": 2830 + }, + { + "epoch": 0.6813819577735125, + "grad_norm": 28.327194690213325, + "learning_rate": 1.3933785339902504e-07, + "logits/chosen": 0.6537632942199707, + "logits/rejected": 0.7060953378677368, + "logps/chosen": -362.89678955078125, + "logps/rejected": -533.9738159179688, + "loss": 0.5279, + "rewards/accuracies": 0.75, + "rewards/chosen": -1.3523266315460205, + "rewards/margins": 1.3014628887176514, + "rewards/rejected": -2.653789520263672, + "step": 2840 + }, + { + "epoch": 0.6837811900191939, + "grad_norm": 30.50433526464709, + "learning_rate": 1.374642233062197e-07, + "logits/chosen": 0.67864990234375, + "logits/rejected": 0.6370582580566406, + "logps/chosen": -456.47265625, + "logps/rejected": -558.3876953125, + "loss": 0.5154, + "rewards/accuracies": 0.800000011920929, + "rewards/chosen": -1.5260423421859741, + "rewards/margins": 1.3706743717193604, + "rewards/rejected": -2.896716594696045, + "step": 2850 + }, + { + "epoch": 0.6861804222648752, + "grad_norm": 34.85452937307168, + "learning_rate": 1.355984871510511e-07, + "logits/chosen": 0.6534871459007263, + "logits/rejected": 0.6843494772911072, + "logps/chosen": -505.02880859375, + "logps/rejected": -617.3344116210938, + "loss": 0.4642, + "rewards/accuracies": 0.8500000238418579, + "rewards/chosen": -1.8383209705352783, + "rewards/margins": 1.1424758434295654, + "rewards/rejected": -2.9807963371276855, + "step": 2860 + }, + { + "epoch": 0.6885796545105566, + "grad_norm": 27.148350418756195, + "learning_rate": 1.3374077580749783e-07, + "logits/chosen": 0.5433071851730347, + "logits/rejected": 0.562809944152832, + "logps/chosen": -321.7220153808594, + "logps/rejected": -465.24078369140625, + "loss": 0.4908, + "rewards/accuracies": 0.7749999761581421, + "rewards/chosen": -1.1553895473480225, + "rewards/margins": 1.265106439590454, + "rewards/rejected": -2.4204962253570557, + "step": 2870 + }, + { + "epoch": 0.690978886756238, + "grad_norm": 46.622796595249234, + "learning_rate": 1.3189121958663024e-07, + "logits/chosen": 0.746674656867981, + "logits/rejected": 0.7814425230026245, + "logps/chosen": -480.2022399902344, + "logps/rejected": -509.8692321777344, + "loss": 0.5062, + "rewards/accuracies": 0.550000011920929, + "rewards/chosen": -1.7685315608978271, + "rewards/margins": 0.6169921159744263, + "rewards/rejected": -2.385523557662964, + "step": 2880 + }, + { + "epoch": 0.6933781190019194, + "grad_norm": 30.590228339686885, + "learning_rate": 1.3004994822746895e-07, + "logits/chosen": 0.5543047189712524, + "logits/rejected": 0.5367203950881958, + "logps/chosen": -393.0322265625, + "logps/rejected": -499.9285583496094, + "loss": 0.5179, + "rewards/accuracies": 0.7250000238418579, + "rewards/chosen": -1.2676035165786743, + "rewards/margins": 1.047071099281311, + "rewards/rejected": -2.3146743774414062, + "step": 2890 + }, + { + "epoch": 0.6957773512476008, + "grad_norm": 32.70639486712309, + "learning_rate": 1.2821709088788434e-07, + "logits/chosen": 0.7046272158622742, + "logits/rejected": 0.7386664152145386, + "logps/chosen": -361.9070739746094, + "logps/rejected": -509.32611083984375, + "loss": 0.5327, + "rewards/accuracies": 0.800000011920929, + "rewards/chosen": -1.3692398071289062, + "rewards/margins": 1.468427300453186, + "rewards/rejected": -2.837667226791382, + "step": 2900 + }, + { + "epoch": 0.6981765834932822, + "grad_norm": 30.843888239231276, + "learning_rate": 1.2639277613553736e-07, + "logits/chosen": 0.7129692435264587, + "logits/rejected": 0.7324257493019104, + "logps/chosen": -339.58807373046875, + "logps/rejected": -462.89324951171875, + "loss": 0.4887, + "rewards/accuracies": 0.7749999761581421, + "rewards/chosen": -1.1596901416778564, + "rewards/margins": 1.1927125453948975, + "rewards/rejected": -2.352402448654175, + "step": 2910 + }, + { + "epoch": 0.7005758157389635, + "grad_norm": 28.751873470088746, + "learning_rate": 1.2457713193885975e-07, + "logits/chosen": 0.5853630304336548, + "logits/rejected": 0.6571609377861023, + "logps/chosen": -337.8079528808594, + "logps/rejected": -537.3619995117188, + "loss": 0.4821, + "rewards/accuracies": 0.7749999761581421, + "rewards/chosen": -1.4434888362884521, + "rewards/margins": 1.6413323879241943, + "rewards/rejected": -3.0848214626312256, + "step": 2920 + }, + { + "epoch": 0.7029750479846449, + "grad_norm": 38.705907816215095, + "learning_rate": 1.2277028565807838e-07, + "logits/chosen": 0.5667808055877686, + "logits/rejected": 0.5656407475471497, + "logps/chosen": -399.50091552734375, + "logps/rejected": -487.841796875, + "loss": 0.5, + "rewards/accuracies": 0.699999988079071, + "rewards/chosen": -1.2812906503677368, + "rewards/margins": 1.0314573049545288, + "rewards/rejected": -2.3127479553222656, + "step": 2930 + }, + { + "epoch": 0.7053742802303263, + "grad_norm": 29.141993488554476, + "learning_rate": 1.209723640362815e-07, + "logits/chosen": 0.5703980326652527, + "logits/rejected": 0.6396486163139343, + "logps/chosen": -431.0738220214844, + "logps/rejected": -601.056884765625, + "loss": 0.5481, + "rewards/accuracies": 0.800000011920929, + "rewards/chosen": -1.4801702499389648, + "rewards/margins": 1.805931806564331, + "rewards/rejected": -3.286102294921875, + "step": 2940 + }, + { + "epoch": 0.7077735124760077, + "grad_norm": 29.731523931470324, + "learning_rate": 1.191834931905277e-07, + "logits/chosen": 0.6887638568878174, + "logits/rejected": 0.6721312403678894, + "logps/chosen": -482.31561279296875, + "logps/rejected": -583.8924560546875, + "loss": 0.5162, + "rewards/accuracies": 0.8500000238418579, + "rewards/chosen": -1.5862760543823242, + "rewards/margins": 1.089349389076233, + "rewards/rejected": -2.6756255626678467, + "step": 2950 + }, + { + "epoch": 0.710172744721689, + "grad_norm": 28.832468635301804, + "learning_rate": 1.1740379860299988e-07, + "logits/chosen": 0.7398797273635864, + "logits/rejected": 0.7511605024337769, + "logps/chosen": -458.5384826660156, + "logps/rejected": -593.3927612304688, + "loss": 0.5, + "rewards/accuracies": 0.7250000238418579, + "rewards/chosen": -1.4789142608642578, + "rewards/margins": 1.1796139478683472, + "rewards/rejected": -2.6585280895233154, + "step": 2960 + }, + { + "epoch": 0.7125719769673704, + "grad_norm": 37.91461771248107, + "learning_rate": 1.1563340511220254e-07, + "logits/chosen": 0.6813124418258667, + "logits/rejected": 0.7123424410820007, + "logps/chosen": -483.406005859375, + "logps/rejected": -592.3176879882812, + "loss": 0.5203, + "rewards/accuracies": 0.699999988079071, + "rewards/chosen": -1.7192113399505615, + "rewards/margins": 1.3098422288894653, + "rewards/rejected": -3.0290539264678955, + "step": 2970 + }, + { + "epoch": 0.7149712092130518, + "grad_norm": 30.53434537820314, + "learning_rate": 1.1387243690420556e-07, + "logits/chosen": 0.7137466669082642, + "logits/rejected": 0.6785651445388794, + "logps/chosen": -489.4556579589844, + "logps/rejected": -642.427490234375, + "loss": 0.4952, + "rewards/accuracies": 0.800000011920929, + "rewards/chosen": -1.4436726570129395, + "rewards/margins": 1.7120336294174194, + "rewards/rejected": -3.1557064056396484, + "step": 2980 + }, + { + "epoch": 0.7173704414587332, + "grad_norm": 46.86299931832656, + "learning_rate": 1.1212101750393235e-07, + "logits/chosen": 0.6887821555137634, + "logits/rejected": 0.7490508556365967, + "logps/chosen": -431.67022705078125, + "logps/rejected": -574.1552734375, + "loss": 0.486, + "rewards/accuracies": 0.8500000238418579, + "rewards/chosen": -1.5646417140960693, + "rewards/margins": 1.6732689142227173, + "rewards/rejected": -3.237910747528076, + "step": 2990 + }, + { + "epoch": 0.7197696737044146, + "grad_norm": 28.83675788521102, + "learning_rate": 1.1037926976649562e-07, + "logits/chosen": 0.6017109751701355, + "logits/rejected": 0.5938320755958557, + "logps/chosen": -454.86676025390625, + "logps/rejected": -602.7736206054688, + "loss": 0.5293, + "rewards/accuracies": 0.7749999761581421, + "rewards/chosen": -1.750307321548462, + "rewards/margins": 1.2305312156677246, + "rewards/rejected": -2.9808382987976074, + "step": 3000 + }, + { + "epoch": 0.722168905950096, + "grad_norm": 39.34411253851995, + "learning_rate": 1.0864731586857936e-07, + "logits/chosen": 0.6451660394668579, + "logits/rejected": 0.6789873838424683, + "logps/chosen": -441.163330078125, + "logps/rejected": -547.3287353515625, + "loss": 0.4773, + "rewards/accuracies": 0.8999999761581421, + "rewards/chosen": -1.3598965406417847, + "rewards/margins": 1.3745959997177124, + "rewards/rejected": -2.734492540359497, + "step": 3010 + }, + { + "epoch": 0.7245681381957774, + "grad_norm": 33.80213582312191, + "learning_rate": 1.0692527729986839e-07, + "logits/chosen": 0.6281944513320923, + "logits/rejected": 0.6428082585334778, + "logps/chosen": -436.121826171875, + "logps/rejected": -549.109619140625, + "loss": 0.4539, + "rewards/accuracies": 0.7749999761581421, + "rewards/chosen": -1.5763047933578491, + "rewards/margins": 1.3722374439239502, + "rewards/rejected": -2.9485421180725098, + "step": 3020 + }, + { + "epoch": 0.7269673704414588, + "grad_norm": 33.8504088811558, + "learning_rate": 1.0521327485452692e-07, + "logits/chosen": 0.6952093839645386, + "logits/rejected": 0.7127727270126343, + "logps/chosen": -426.577880859375, + "logps/rejected": -585.4749145507812, + "loss": 0.4779, + "rewards/accuracies": 0.7749999761581421, + "rewards/chosen": -1.641758680343628, + "rewards/margins": 1.7743602991104126, + "rewards/rejected": -3.41611909866333, + "step": 3030 + }, + { + "epoch": 0.7293666026871402, + "grad_norm": 36.51412890934737, + "learning_rate": 1.0351142862272468e-07, + "logits/chosen": 0.7390758395195007, + "logits/rejected": 0.7217782139778137, + "logps/chosen": -395.6050720214844, + "logps/rejected": -596.3590087890625, + "loss": 0.506, + "rewards/accuracies": 0.8500000238418579, + "rewards/chosen": -1.6227900981903076, + "rewards/margins": 2.1645493507385254, + "rewards/rejected": -3.787339448928833, + "step": 3040 + }, + { + "epoch": 0.7317658349328215, + "grad_norm": 30.15398627394634, + "learning_rate": 1.0181985798221343e-07, + "logits/chosen": 0.6722254753112793, + "logits/rejected": 0.7073189616203308, + "logps/chosen": -420.848876953125, + "logps/rejected": -572.2374877929688, + "loss": 0.5061, + "rewards/accuracies": 0.875, + "rewards/chosen": -1.448339819908142, + "rewards/margins": 1.4826971292495728, + "rewards/rejected": -2.931036949157715, + "step": 3050 + }, + { + "epoch": 0.7341650671785028, + "grad_norm": 33.15513048467062, + "learning_rate": 1.0013868158995329e-07, + "logits/chosen": 0.8221071362495422, + "logits/rejected": 0.8116429448127747, + "logps/chosen": -450.3140563964844, + "logps/rejected": -565.424072265625, + "loss": 0.4946, + "rewards/accuracies": 0.824999988079071, + "rewards/chosen": -1.6256906986236572, + "rewards/margins": 1.3659015893936157, + "rewards/rejected": -2.9915921688079834, + "step": 3060 + }, + { + "epoch": 0.7365642994241842, + "grad_norm": 28.899348484638733, + "learning_rate": 9.84680173737887e-08, + "logits/chosen": 0.7205225825309753, + "logits/rejected": 0.6872090101242065, + "logps/chosen": -433.659912109375, + "logps/rejected": -501.6417541503906, + "loss": 0.5089, + "rewards/accuracies": 0.7250000238418579, + "rewards/chosen": -1.4491467475891113, + "rewards/margins": 1.1040198802947998, + "rewards/rejected": -2.553166389465332, + "step": 3070 + }, + { + "epoch": 0.7389635316698656, + "grad_norm": 31.437945405932854, + "learning_rate": 9.680798252417713e-08, + "logits/chosen": 0.7566556930541992, + "logits/rejected": 0.7423285245895386, + "logps/chosen": -362.8130798339844, + "logps/rejected": -492.66058349609375, + "loss": 0.4928, + "rewards/accuracies": 0.8500000238418579, + "rewards/chosen": -1.347149133682251, + "rewards/margins": 0.9845093488693237, + "rewards/rejected": -2.3316586017608643, + "step": 3080 + }, + { + "epoch": 0.741362763915547, + "grad_norm": 28.49317814535035, + "learning_rate": 9.515869348596808e-08, + "logits/chosen": 0.7042705416679382, + "logits/rejected": 0.6866374611854553, + "logps/chosen": -441.2322692871094, + "logps/rejected": -548.809814453125, + "loss": 0.476, + "rewards/accuracies": 0.875, + "rewards/chosen": -1.3135347366333008, + "rewards/margins": 1.3202944993972778, + "rewards/rejected": -2.6338295936584473, + "step": 3090 + }, + { + "epoch": 0.7437619961612284, + "grad_norm": 44.16067627194287, + "learning_rate": 9.352026595023493e-08, + "logits/chosen": 0.6853736042976379, + "logits/rejected": 0.7283433675765991, + "logps/chosen": -435.65362548828125, + "logps/rejected": -469.2335510253906, + "loss": 0.4944, + "rewards/accuracies": 0.6499999761581421, + "rewards/chosen": -1.3361632823944092, + "rewards/margins": 0.6489462852478027, + "rewards/rejected": -1.9851096868515015, + "step": 3100 + }, + { + "epoch": 0.7461612284069098, + "grad_norm": 39.47362796340898, + "learning_rate": 9.189281484616004e-08, + "logits/chosen": 0.7125884294509888, + "logits/rejected": 0.7073049545288086, + "logps/chosen": -366.3876953125, + "logps/rejected": -517.1468505859375, + "loss": 0.5364, + "rewards/accuracies": 0.699999988079071, + "rewards/chosen": -1.4312505722045898, + "rewards/margins": 1.1074860095977783, + "rewards/rejected": -2.538736343383789, + "step": 3110 + }, + { + "epoch": 0.7485604606525912, + "grad_norm": 44.465462304398336, + "learning_rate": 9.027645433297249e-08, + "logits/chosen": 0.6530724167823792, + "logits/rejected": 0.6677568554878235, + "logps/chosen": -544.2484130859375, + "logps/rejected": -627.7911376953125, + "loss": 0.5358, + "rewards/accuracies": 0.7749999761581421, + "rewards/chosen": -2.0512328147888184, + "rewards/margins": 1.2890361547470093, + "rewards/rejected": -3.340269088745117, + "step": 3120 + }, + { + "epoch": 0.7509596928982726, + "grad_norm": 41.65465122772129, + "learning_rate": 8.867129779194066e-08, + "logits/chosen": 0.6696042418479919, + "logits/rejected": 0.6906247138977051, + "logps/chosen": -351.27984619140625, + "logps/rejected": -566.343017578125, + "loss": 0.5074, + "rewards/accuracies": 0.7250000238418579, + "rewards/chosen": -1.212708830833435, + "rewards/margins": 2.181748628616333, + "rewards/rejected": -3.3944575786590576, + "step": 3130 + }, + { + "epoch": 0.753358925143954, + "grad_norm": 37.16565124609395, + "learning_rate": 8.707745781841866e-08, + "logits/chosen": 0.7126466035842896, + "logits/rejected": 0.6710564494132996, + "logps/chosen": -423.04034423828125, + "logps/rejected": -557.0213623046875, + "loss": 0.4971, + "rewards/accuracies": 0.824999988079071, + "rewards/chosen": -1.7721456289291382, + "rewards/margins": 1.449415922164917, + "rewards/rejected": -3.2215614318847656, + "step": 3140 + }, + { + "epoch": 0.7557581573896354, + "grad_norm": 23.599726604923614, + "learning_rate": 8.549504621394831e-08, + "logits/chosen": 0.7398973703384399, + "logits/rejected": 0.751029372215271, + "logps/chosen": -400.4888000488281, + "logps/rejected": -586.410400390625, + "loss": 0.4405, + "rewards/accuracies": 0.75, + "rewards/chosen": -1.4728138446807861, + "rewards/margins": 1.921562910079956, + "rewards/rejected": -3.394376754760742, + "step": 3150 + }, + { + "epoch": 0.7581573896353166, + "grad_norm": 47.47497023724174, + "learning_rate": 8.392417397841703e-08, + "logits/chosen": 0.7788910269737244, + "logits/rejected": 0.8002176284790039, + "logps/chosen": -407.0137634277344, + "logps/rejected": -546.9268798828125, + "loss": 0.4876, + "rewards/accuracies": 0.8500000238418579, + "rewards/chosen": -1.223954439163208, + "rewards/margins": 1.2972733974456787, + "rewards/rejected": -2.5212275981903076, + "step": 3160 + }, + { + "epoch": 0.760556621880998, + "grad_norm": 54.01121345632537, + "learning_rate": 8.236495130227083e-08, + "logits/chosen": 0.6500368714332581, + "logits/rejected": 0.7030578851699829, + "logps/chosen": -501.3116149902344, + "logps/rejected": -648.521728515625, + "loss": 0.516, + "rewards/accuracies": 0.824999988079071, + "rewards/chosen": -1.90944504737854, + "rewards/margins": 1.923628568649292, + "rewards/rejected": -3.833073377609253, + "step": 3170 + }, + { + "epoch": 0.7629558541266794, + "grad_norm": 36.20984331034919, + "learning_rate": 8.081748755878612e-08, + "logits/chosen": 0.7043317556381226, + "logits/rejected": 0.7164947986602783, + "logps/chosen": -452.80145263671875, + "logps/rejected": -507.310791015625, + "loss": 0.4931, + "rewards/accuracies": 0.824999988079071, + "rewards/chosen": -1.5479075908660889, + "rewards/margins": 1.1401413679122925, + "rewards/rejected": -2.688048839569092, + "step": 3180 + }, + { + "epoch": 0.7653550863723608, + "grad_norm": 32.72403763285082, + "learning_rate": 7.928189129639632e-08, + "logits/chosen": 0.7559349536895752, + "logits/rejected": 0.7914206385612488, + "logps/chosen": -390.57232666015625, + "logps/rejected": -523.0045776367188, + "loss": 0.4761, + "rewards/accuracies": 0.875, + "rewards/chosen": -1.420803427696228, + "rewards/margins": 1.2412652969360352, + "rewards/rejected": -2.6620688438415527, + "step": 3190 + }, + { + "epoch": 0.7677543186180422, + "grad_norm": 41.93143711138968, + "learning_rate": 7.775827023107834e-08, + "logits/chosen": 0.7086448073387146, + "logits/rejected": 0.6970892548561096, + "logps/chosen": -408.8316345214844, + "logps/rejected": -539.7716064453125, + "loss": 0.5404, + "rewards/accuracies": 0.699999988079071, + "rewards/chosen": -1.546519160270691, + "rewards/margins": 1.1031622886657715, + "rewards/rejected": -2.649681568145752, + "step": 3200 + }, + { + "epoch": 0.7701535508637236, + "grad_norm": 43.87698418658202, + "learning_rate": 7.624673123879682e-08, + "logits/chosen": 0.678581714630127, + "logits/rejected": 0.6445807814598083, + "logps/chosen": -392.3583984375, + "logps/rejected": -468.27911376953125, + "loss": 0.521, + "rewards/accuracies": 0.7749999761581421, + "rewards/chosen": -1.335402488708496, + "rewards/margins": 0.9563083648681641, + "rewards/rejected": -2.29171085357666, + "step": 3210 + }, + { + "epoch": 0.772552783109405, + "grad_norm": 28.844880569875237, + "learning_rate": 7.474738034800663e-08, + "logits/chosen": 0.6901575326919556, + "logits/rejected": 0.7580747008323669, + "logps/chosen": -356.3608093261719, + "logps/rejected": -538.3591918945312, + "loss": 0.5064, + "rewards/accuracies": 0.75, + "rewards/chosen": -1.2393145561218262, + "rewards/margins": 2.0787644386291504, + "rewards/rejected": -3.3180789947509766, + "step": 3220 + }, + { + "epoch": 0.7749520153550864, + "grad_norm": 31.987174559648025, + "learning_rate": 7.326032273221606e-08, + "logits/chosen": 0.6704432368278503, + "logits/rejected": 0.8194705247879028, + "logps/chosen": -471.6807556152344, + "logps/rejected": -564.5304565429688, + "loss": 0.4858, + "rewards/accuracies": 0.7749999761581421, + "rewards/chosen": -1.6247665882110596, + "rewards/margins": 1.192398190498352, + "rewards/rejected": -2.817164659500122, + "step": 3230 + }, + { + "epoch": 0.7773512476007678, + "grad_norm": 33.470079538599556, + "learning_rate": 7.178566270260872e-08, + "logits/chosen": 0.6915429830551147, + "logits/rejected": 0.7422800660133362, + "logps/chosen": -452.6827087402344, + "logps/rejected": -586.0211181640625, + "loss": 0.518, + "rewards/accuracies": 0.625, + "rewards/chosen": -1.7043472528457642, + "rewards/margins": 1.16866934299469, + "rewards/rejected": -2.873016357421875, + "step": 3240 + }, + { + "epoch": 0.7797504798464492, + "grad_norm": 28.19482569283347, + "learning_rate": 7.032350370072709e-08, + "logits/chosen": 0.6776562929153442, + "logits/rejected": 0.7009492516517639, + "logps/chosen": -423.5204162597656, + "logps/rejected": -529.0217895507812, + "loss": 0.4666, + "rewards/accuracies": 0.800000011920929, + "rewards/chosen": -1.3196834325790405, + "rewards/margins": 1.1787526607513428, + "rewards/rejected": -2.4984359741210938, + "step": 3250 + }, + { + "epoch": 0.7821497120921305, + "grad_norm": 25.36749286718069, + "learning_rate": 6.887394829121596e-08, + "logits/chosen": 0.7724124789237976, + "logits/rejected": 0.7605770230293274, + "logps/chosen": -458.8877868652344, + "logps/rejected": -655.66552734375, + "loss": 0.4933, + "rewards/accuracies": 0.800000011920929, + "rewards/chosen": -1.7062854766845703, + "rewards/margins": 2.266158103942871, + "rewards/rejected": -3.9724433422088623, + "step": 3260 + }, + { + "epoch": 0.7845489443378119, + "grad_norm": 31.238032068852384, + "learning_rate": 6.743709815462833e-08, + "logits/chosen": 0.6670851707458496, + "logits/rejected": 0.7134937047958374, + "logps/chosen": -439.99688720703125, + "logps/rejected": -536.1776123046875, + "loss": 0.4783, + "rewards/accuracies": 0.75, + "rewards/chosen": -1.5214483737945557, + "rewards/margins": 1.321410059928894, + "rewards/rejected": -2.8428585529327393, + "step": 3270 + }, + { + "epoch": 0.7869481765834933, + "grad_norm": 33.34150808500615, + "learning_rate": 6.601305408029287e-08, + "logits/chosen": 0.7619131803512573, + "logits/rejected": 0.789661705493927, + "logps/chosen": -435.0225524902344, + "logps/rejected": -573.7952880859375, + "loss": 0.4747, + "rewards/accuracies": 0.800000011920929, + "rewards/chosen": -1.7262579202651978, + "rewards/margins": 1.4246705770492554, + "rewards/rejected": -3.150928258895874, + "step": 3280 + }, + { + "epoch": 0.7893474088291746, + "grad_norm": 29.44124572086776, + "learning_rate": 6.460191595924366e-08, + "logits/chosen": 0.657812774181366, + "logits/rejected": 0.7025941014289856, + "logps/chosen": -416.65948486328125, + "logps/rejected": -534.0513305664062, + "loss": 0.4573, + "rewards/accuracies": 0.7250000238418579, + "rewards/chosen": -1.4878287315368652, + "rewards/margins": 1.1736620664596558, + "rewards/rejected": -2.6614909172058105, + "step": 3290 + }, + { + "epoch": 0.791746641074856, + "grad_norm": 38.40279226701135, + "learning_rate": 6.320378277721342e-08, + "logits/chosen": 0.7255043387413025, + "logits/rejected": 0.7269566655158997, + "logps/chosen": -447.64080810546875, + "logps/rejected": -507.58966064453125, + "loss": 0.4761, + "rewards/accuracies": 0.7250000238418579, + "rewards/chosen": -1.7017924785614014, + "rewards/margins": 0.7691752910614014, + "rewards/rejected": -2.4709675312042236, + "step": 3300 + }, + { + "epoch": 0.7941458733205374, + "grad_norm": 39.684598268718624, + "learning_rate": 6.181875260769032e-08, + "logits/chosen": 0.6789619326591492, + "logits/rejected": 0.6977161169052124, + "logps/chosen": -454.3805236816406, + "logps/rejected": -490.30908203125, + "loss": 0.4726, + "rewards/accuracies": 0.7749999761581421, + "rewards/chosen": -1.4421659708023071, + "rewards/margins": 1.0759384632110596, + "rewards/rejected": -2.5181047916412354, + "step": 3310 + }, + { + "epoch": 0.7965451055662188, + "grad_norm": 34.83351031389863, + "learning_rate": 6.044692260503797e-08, + "logits/chosen": 0.768031895160675, + "logits/rejected": 0.7473145723342896, + "logps/chosen": -501.7557678222656, + "logps/rejected": -619.917236328125, + "loss": 0.4669, + "rewards/accuracies": 0.800000011920929, + "rewards/chosen": -1.8463929891586304, + "rewards/margins": 1.5131759643554688, + "rewards/rejected": -3.3595688343048096, + "step": 3320 + }, + { + "epoch": 0.7989443378119002, + "grad_norm": 35.31190929686924, + "learning_rate": 5.9088388997680984e-08, + "logits/chosen": 0.6237305998802185, + "logits/rejected": 0.6354110240936279, + "logps/chosen": -510.52801513671875, + "logps/rejected": -576.4329833984375, + "loss": 0.4755, + "rewards/accuracies": 0.824999988079071, + "rewards/chosen": -1.5719668865203857, + "rewards/margins": 1.4724681377410889, + "rewards/rejected": -3.0444350242614746, + "step": 3330 + }, + { + "epoch": 0.8013435700575816, + "grad_norm": 41.916022771150374, + "learning_rate": 5.774324708135439e-08, + "logits/chosen": 0.7030372619628906, + "logits/rejected": 0.719199538230896, + "logps/chosen": -357.8835144042969, + "logps/rejected": -461.7693786621094, + "loss": 0.49, + "rewards/accuracies": 0.7749999761581421, + "rewards/chosen": -1.2769114971160889, + "rewards/margins": 1.2639166116714478, + "rewards/rejected": -2.540827751159668, + "step": 3340 + }, + { + "epoch": 0.803742802303263, + "grad_norm": 27.73526705705633, + "learning_rate": 5.641159121241953e-08, + "logits/chosen": 0.7158384919166565, + "logits/rejected": 0.7426877021789551, + "logps/chosen": -401.23834228515625, + "logps/rejected": -585.4378051757812, + "loss": 0.4903, + "rewards/accuracies": 0.75, + "rewards/chosen": -1.5427526235580444, + "rewards/margins": 1.556259274482727, + "rewards/rejected": -3.0990118980407715, + "step": 3350 + }, + { + "epoch": 0.8061420345489443, + "grad_norm": 44.61122166912319, + "learning_rate": 5.5093514801245106e-08, + "logits/chosen": 0.6555548310279846, + "logits/rejected": 0.6843029260635376, + "logps/chosen": -425.64776611328125, + "logps/rejected": -592.9639892578125, + "loss": 0.5154, + "rewards/accuracies": 0.7749999761581421, + "rewards/chosen": -1.610778570175171, + "rewards/margins": 1.449095606803894, + "rewards/rejected": -3.0598740577697754, + "step": 3360 + }, + { + "epoch": 0.8085412667946257, + "grad_norm": 27.134645439655188, + "learning_rate": 5.378911030565453e-08, + "logits/chosen": 0.650478184223175, + "logits/rejected": 0.6665130257606506, + "logps/chosen": -483.04779052734375, + "logps/rejected": -646.669677734375, + "loss": 0.4583, + "rewards/accuracies": 0.824999988079071, + "rewards/chosen": -1.76322340965271, + "rewards/margins": 1.3918665647506714, + "rewards/rejected": -3.155089855194092, + "step": 3370 + }, + { + "epoch": 0.8109404990403071, + "grad_norm": 30.88613336840347, + "learning_rate": 5.249846922444101e-08, + "logits/chosen": 0.6785479784011841, + "logits/rejected": 0.7193585634231567, + "logps/chosen": -427.58758544921875, + "logps/rejected": -651.4105224609375, + "loss": 0.48, + "rewards/accuracies": 0.7250000238418579, + "rewards/chosen": -1.9091869592666626, + "rewards/margins": 2.4718122482299805, + "rewards/rejected": -4.380999565124512, + "step": 3380 + }, + { + "epoch": 0.8133397312859885, + "grad_norm": 39.208213968014256, + "learning_rate": 5.122168209094865e-08, + "logits/chosen": 0.6814306378364563, + "logits/rejected": 0.7048656344413757, + "logps/chosen": -378.8595886230469, + "logps/rejected": -436.8194274902344, + "loss": 0.5038, + "rewards/accuracies": 0.7250000238418579, + "rewards/chosen": -1.463390588760376, + "rewards/margins": 0.6034170985221863, + "rewards/rejected": -2.066807746887207, + "step": 3390 + }, + { + "epoch": 0.8157389635316699, + "grad_norm": 28.560854060893856, + "learning_rate": 4.995883846672222e-08, + "logits/chosen": 0.6933716535568237, + "logits/rejected": 0.6919404864311218, + "logps/chosen": -597.2493896484375, + "logps/rejected": -596.7833251953125, + "loss": 0.4986, + "rewards/accuracies": 0.7250000238418579, + "rewards/chosen": -2.024073600769043, + "rewards/margins": 0.7939941883087158, + "rewards/rejected": -2.818067789077759, + "step": 3400 + }, + { + "epoch": 0.8181381957773513, + "grad_norm": 31.695811688746335, + "learning_rate": 4.871002693522486e-08, + "logits/chosen": 0.6247652173042297, + "logits/rejected": 0.6395012736320496, + "logps/chosen": -460.6663513183594, + "logps/rejected": -528.9644775390625, + "loss": 0.4888, + "rewards/accuracies": 0.7749999761581421, + "rewards/chosen": -1.649450659751892, + "rewards/margins": 1.0935943126678467, + "rewards/rejected": -2.7430450916290283, + "step": 3410 + }, + { + "epoch": 0.8205374280230326, + "grad_norm": 31.093359172893916, + "learning_rate": 4.7475335095623956e-08, + "logits/chosen": 0.7312600612640381, + "logits/rejected": 0.7648332715034485, + "logps/chosen": -457.8077697753906, + "logps/rejected": -568.5828247070312, + "loss": 0.4947, + "rewards/accuracies": 0.7250000238418579, + "rewards/chosen": -1.7661203145980835, + "rewards/margins": 1.2773500680923462, + "rewards/rejected": -3.0434701442718506, + "step": 3420 + }, + { + "epoch": 0.822936660268714, + "grad_norm": 51.951004563036165, + "learning_rate": 4.6254849556646714e-08, + "logits/chosen": 0.5918694734573364, + "logits/rejected": 0.6510140299797058, + "logps/chosen": -518.0965576171875, + "logps/rejected": -644.2404174804688, + "loss": 0.5116, + "rewards/accuracies": 0.699999988079071, + "rewards/chosen": -1.8913600444793701, + "rewards/margins": 1.7554857730865479, + "rewards/rejected": -3.646846055984497, + "step": 3430 + }, + { + "epoch": 0.8253358925143954, + "grad_norm": 30.668402411962806, + "learning_rate": 4.504865593050483e-08, + "logits/chosen": 0.6927034258842468, + "logits/rejected": 0.7566369771957397, + "logps/chosen": -466.992431640625, + "logps/rejected": -573.04638671875, + "loss": 0.5, + "rewards/accuracies": 0.699999988079071, + "rewards/chosen": -1.7490732669830322, + "rewards/margins": 1.085127353668213, + "rewards/rejected": -2.834200382232666, + "step": 3440 + }, + { + "epoch": 0.8277351247600768, + "grad_norm": 33.39654631864919, + "learning_rate": 4.385683882688895e-08, + "logits/chosen": 0.6728897094726562, + "logits/rejected": 0.7135699987411499, + "logps/chosen": -461.8128967285156, + "logps/rejected": -448.58978271484375, + "loss": 0.547, + "rewards/accuracies": 0.6499999761581421, + "rewards/chosen": -1.641575813293457, + "rewards/margins": 0.3381771445274353, + "rewards/rejected": -1.9797531366348267, + "step": 3450 + }, + { + "epoch": 0.8301343570057581, + "grad_norm": 54.916707180210906, + "learning_rate": 4.2679481847033985e-08, + "logits/chosen": 0.7350330352783203, + "logits/rejected": 0.7353192567825317, + "logps/chosen": -449.59814453125, + "logps/rejected": -616.7791137695312, + "loss": 0.5093, + "rewards/accuracies": 0.7250000238418579, + "rewards/chosen": -1.633469820022583, + "rewards/margins": 1.7337658405303955, + "rewards/rejected": -3.3672356605529785, + "step": 3460 + }, + { + "epoch": 0.8325335892514395, + "grad_norm": 34.40985347014142, + "learning_rate": 4.151666757785435e-08, + "logits/chosen": 0.70672607421875, + "logits/rejected": 0.7129568457603455, + "logps/chosen": -392.37969970703125, + "logps/rejected": -576.6778564453125, + "loss": 0.4673, + "rewards/accuracies": 0.8999999761581421, + "rewards/chosen": -1.2961803674697876, + "rewards/margins": 1.9318246841430664, + "rewards/rejected": -3.2280051708221436, + "step": 3470 + }, + { + "epoch": 0.8349328214971209, + "grad_norm": 34.150066441594575, + "learning_rate": 4.036847758615136e-08, + "logits/chosen": 0.7242936491966248, + "logits/rejected": 0.7108465433120728, + "logps/chosen": -474.98565673828125, + "logps/rejected": -555.7901000976562, + "loss": 0.5396, + "rewards/accuracies": 0.550000011920929, + "rewards/chosen": -2.126871109008789, + "rewards/margins": 0.8020073175430298, + "rewards/rejected": -2.9288783073425293, + "step": 3480 + }, + { + "epoch": 0.8373320537428023, + "grad_norm": 34.55135904008657, + "learning_rate": 3.923499241289113e-08, + "logits/chosen": 0.6907113194465637, + "logits/rejected": 0.6848424673080444, + "logps/chosen": -533.1072998046875, + "logps/rejected": -573.22509765625, + "loss": 0.5329, + "rewards/accuracies": 0.675000011920929, + "rewards/chosen": -2.085261106491089, + "rewards/margins": 1.0348024368286133, + "rewards/rejected": -3.120063304901123, + "step": 3490 + }, + { + "epoch": 0.8397312859884837, + "grad_norm": 26.808282974911915, + "learning_rate": 3.811629156755541e-08, + "logits/chosen": 0.6466808319091797, + "logits/rejected": 0.6538549661636353, + "logps/chosen": -456.07049560546875, + "logps/rejected": -551.8894653320312, + "loss": 0.4774, + "rewards/accuracies": 0.75, + "rewards/chosen": -1.423441767692566, + "rewards/margins": 1.1218823194503784, + "rewards/rejected": -2.5453243255615234, + "step": 3500 + }, + { + "epoch": 0.8421305182341651, + "grad_norm": 28.83291099893329, + "learning_rate": 3.701245352256391e-08, + "logits/chosen": 0.6636977195739746, + "logits/rejected": 0.6604019999504089, + "logps/chosen": -456.4715270996094, + "logps/rejected": -504.9053649902344, + "loss": 0.4905, + "rewards/accuracies": 0.800000011920929, + "rewards/chosen": -1.4351160526275635, + "rewards/margins": 0.7614234685897827, + "rewards/rejected": -2.1965396404266357, + "step": 3510 + }, + { + "epoch": 0.8445297504798465, + "grad_norm": 42.02738859118274, + "learning_rate": 3.592355570776984e-08, + "logits/chosen": 0.6333275437355042, + "logits/rejected": 0.6459077596664429, + "logps/chosen": -352.7582092285156, + "logps/rejected": -492.6253967285156, + "loss": 0.4725, + "rewards/accuracies": 0.8500000238418579, + "rewards/chosen": -1.1113379001617432, + "rewards/margins": 1.3925281763076782, + "rewards/rejected": -2.503865957260132, + "step": 3520 + }, + { + "epoch": 0.8469289827255279, + "grad_norm": 28.592815771886347, + "learning_rate": 3.484967450502904e-08, + "logits/chosen": 0.8196449279785156, + "logits/rejected": 0.7500802874565125, + "logps/chosen": -360.6011657714844, + "logps/rejected": -559.92919921875, + "loss": 0.4666, + "rewards/accuracies": 0.800000011920929, + "rewards/chosen": -1.395541787147522, + "rewards/margins": 1.602349042892456, + "rewards/rejected": -2.9978909492492676, + "step": 3530 + }, + { + "epoch": 0.8493282149712092, + "grad_norm": 52.0300622408455, + "learning_rate": 3.3790885242841296e-08, + "logits/chosen": 0.6689568758010864, + "logits/rejected": 0.7228366136550903, + "logps/chosen": -469.3619079589844, + "logps/rejected": -608.2591552734375, + "loss": 0.4778, + "rewards/accuracies": 0.7250000238418579, + "rewards/chosen": -1.9589526653289795, + "rewards/margins": 1.4972124099731445, + "rewards/rejected": -3.456165313720703, + "step": 3540 + }, + { + "epoch": 0.8517274472168906, + "grad_norm": 44.27630823570909, + "learning_rate": 3.274726219106677e-08, + "logits/chosen": 0.7316599488258362, + "logits/rejected": 0.756661593914032, + "logps/chosen": -514.4577026367188, + "logps/rejected": -639.2863159179688, + "loss": 0.4963, + "rewards/accuracies": 0.7749999761581421, + "rewards/chosen": -2.047656297683716, + "rewards/margins": 1.3594030141830444, + "rewards/rejected": -3.407059907913208, + "step": 3550 + }, + { + "epoch": 0.8541266794625719, + "grad_norm": 28.64271031691096, + "learning_rate": 3.171887855571642e-08, + "logits/chosen": 0.6066162586212158, + "logits/rejected": 0.6715503931045532, + "logps/chosen": -382.43878173828125, + "logps/rejected": -457.4537048339844, + "loss": 0.4803, + "rewards/accuracies": 0.675000011920929, + "rewards/chosen": -1.4242990016937256, + "rewards/margins": 0.8341078758239746, + "rewards/rejected": -2.2584068775177, + "step": 3560 + }, + { + "epoch": 0.8565259117082533, + "grad_norm": 51.494044930017054, + "learning_rate": 3.070580647381643e-08, + "logits/chosen": 0.7110476493835449, + "logits/rejected": 0.6737565994262695, + "logps/chosen": -428.01654052734375, + "logps/rejected": -599.3654174804688, + "loss": 0.4971, + "rewards/accuracies": 0.824999988079071, + "rewards/chosen": -1.751559853553772, + "rewards/margins": 1.7291587591171265, + "rewards/rejected": -3.4807181358337402, + "step": 3570 + }, + { + "epoch": 0.8589251439539347, + "grad_norm": 32.8193357126353, + "learning_rate": 2.9708117008348576e-08, + "logits/chosen": 0.8528485298156738, + "logits/rejected": 0.8270418047904968, + "logps/chosen": -485.8948669433594, + "logps/rejected": -512.58935546875, + "loss": 0.4711, + "rewards/accuracies": 0.699999988079071, + "rewards/chosen": -1.5045112371444702, + "rewards/margins": 0.845577597618103, + "rewards/rejected": -2.3500890731811523, + "step": 3580 + }, + { + "epoch": 0.8613243761996161, + "grad_norm": 42.77177983927304, + "learning_rate": 2.8725880143264992e-08, + "logits/chosen": 0.6248888373374939, + "logits/rejected": 0.6356110572814941, + "logps/chosen": -448.6212463378906, + "logps/rejected": -572.7357177734375, + "loss": 0.522, + "rewards/accuracies": 0.75, + "rewards/chosen": -1.826633095741272, + "rewards/margins": 0.8497768640518188, + "rewards/rejected": -2.67641019821167, + "step": 3590 + }, + { + "epoch": 0.8637236084452975, + "grad_norm": 45.77767199385067, + "learning_rate": 2.775916477857948e-08, + "logits/chosen": 0.7616924047470093, + "logits/rejected": 0.782433271408081, + "logps/chosen": -378.56072998046875, + "logps/rejected": -471.501708984375, + "loss": 0.4733, + "rewards/accuracies": 0.7749999761581421, + "rewards/chosen": -1.4830647706985474, + "rewards/margins": 0.9338682293891907, + "rewards/rejected": -2.416933059692383, + "step": 3600 + }, + { + "epoch": 0.8661228406909789, + "grad_norm": 38.72474554060603, + "learning_rate": 2.680803872553408e-08, + "logits/chosen": 0.6768021583557129, + "logits/rejected": 0.6709715723991394, + "logps/chosen": -418.63922119140625, + "logps/rejected": -630.6937255859375, + "loss": 0.4942, + "rewards/accuracies": 0.8500000238418579, + "rewards/chosen": -1.439439296722412, + "rewards/margins": 2.4228386878967285, + "rewards/rejected": -3.8622779846191406, + "step": 3610 + }, + { + "epoch": 0.8685220729366603, + "grad_norm": 57.14033737148358, + "learning_rate": 2.5872568701842706e-08, + "logits/chosen": 0.6473024487495422, + "logits/rejected": 0.6710727214813232, + "logps/chosen": -385.234375, + "logps/rejected": -537.0969848632812, + "loss": 0.5306, + "rewards/accuracies": 0.675000011920929, + "rewards/chosen": -1.539281964302063, + "rewards/margins": 1.319455862045288, + "rewards/rejected": -2.8587379455566406, + "step": 3620 + }, + { + "epoch": 0.8709213051823417, + "grad_norm": 42.126294732393575, + "learning_rate": 2.495282032701096e-08, + "logits/chosen": 0.7393206357955933, + "logits/rejected": 0.7721438407897949, + "logps/chosen": -342.84613037109375, + "logps/rejected": -438.0054626464844, + "loss": 0.4718, + "rewards/accuracies": 0.7749999761581421, + "rewards/chosen": -1.2897088527679443, + "rewards/margins": 1.3247145414352417, + "rewards/rejected": -2.6144232749938965, + "step": 3630 + }, + { + "epoch": 0.8733205374280231, + "grad_norm": 45.48215970360758, + "learning_rate": 2.4048858117733133e-08, + "logits/chosen": 0.6715846061706543, + "logits/rejected": 0.7098231911659241, + "logps/chosen": -445.4051818847656, + "logps/rejected": -614.8271484375, + "loss": 0.4511, + "rewards/accuracies": 0.7250000238418579, + "rewards/chosen": -1.5650415420532227, + "rewards/margins": 2.225489377975464, + "rewards/rejected": -3.7905311584472656, + "step": 3640 + }, + { + "epoch": 0.8757197696737045, + "grad_norm": 33.58800202447882, + "learning_rate": 2.3160745483366938e-08, + "logits/chosen": 0.6484588980674744, + "logits/rejected": 0.6462217569351196, + "logps/chosen": -418.74310302734375, + "logps/rejected": -575.1314086914062, + "loss": 0.467, + "rewards/accuracies": 0.8500000238418579, + "rewards/chosen": -1.5988870859146118, + "rewards/margins": 1.3143718242645264, + "rewards/rejected": -2.9132585525512695, + "step": 3650 + }, + { + "epoch": 0.8781190019193857, + "grad_norm": 35.56501086040675, + "learning_rate": 2.2288544721485197e-08, + "logits/chosen": 0.6811990141868591, + "logits/rejected": 0.6851666569709778, + "logps/chosen": -363.73486328125, + "logps/rejected": -558.9583740234375, + "loss": 0.464, + "rewards/accuracies": 0.800000011920929, + "rewards/chosen": -1.2833194732666016, + "rewards/margins": 1.7774661779403687, + "rewards/rejected": -3.0607855319976807, + "step": 3660 + }, + { + "epoch": 0.8805182341650671, + "grad_norm": 48.885323582555905, + "learning_rate": 2.1432317013506117e-08, + "logits/chosen": 0.6057692766189575, + "logits/rejected": 0.6347171664237976, + "logps/chosen": -456.45050048828125, + "logps/rejected": -491.285888671875, + "loss": 0.5064, + "rewards/accuracies": 0.699999988079071, + "rewards/chosen": -1.7665584087371826, + "rewards/margins": 0.7946644425392151, + "rewards/rejected": -2.561223268508911, + "step": 3670 + }, + { + "epoch": 0.8829174664107485, + "grad_norm": 52.026359398966434, + "learning_rate": 2.0592122420401704e-08, + "logits/chosen": 0.68748939037323, + "logits/rejected": 0.7001315355300903, + "logps/chosen": -417.6197204589844, + "logps/rejected": -501.4674377441406, + "loss": 0.523, + "rewards/accuracies": 0.75, + "rewards/chosen": -1.6828155517578125, + "rewards/margins": 0.8300768733024597, + "rewards/rejected": -2.512892246246338, + "step": 3680 + }, + { + "epoch": 0.8853166986564299, + "grad_norm": 40.657536973128096, + "learning_rate": 1.976801987848459e-08, + "logits/chosen": 0.7110599279403687, + "logits/rejected": 0.7387591600418091, + "logps/chosen": -483.6581115722656, + "logps/rejected": -655.469970703125, + "loss": 0.509, + "rewards/accuracies": 0.824999988079071, + "rewards/chosen": -1.9261047840118408, + "rewards/margins": 1.5848186016082764, + "rewards/rejected": -3.510922908782959, + "step": 3690 + }, + { + "epoch": 0.8877159309021113, + "grad_norm": 43.24860595798503, + "learning_rate": 1.8960067195273987e-08, + "logits/chosen": 0.6262224316596985, + "logits/rejected": 0.6679535508155823, + "logps/chosen": -383.0298156738281, + "logps/rejected": -520.960205078125, + "loss": 0.5095, + "rewards/accuracies": 0.75, + "rewards/chosen": -1.4945952892303467, + "rewards/margins": 1.4201505184173584, + "rewards/rejected": -2.914746046066284, + "step": 3700 + }, + { + "epoch": 0.8901151631477927, + "grad_norm": 30.03259002192758, + "learning_rate": 1.816832104544072e-08, + "logits/chosen": 0.7243183851242065, + "logits/rejected": 0.7136281132698059, + "logps/chosen": -464.9974060058594, + "logps/rejected": -572.7579956054688, + "loss": 0.4569, + "rewards/accuracies": 0.7749999761581421, + "rewards/chosen": -1.7093970775604248, + "rewards/margins": 1.3652998208999634, + "rewards/rejected": -3.0746970176696777, + "step": 3710 + }, + { + "epoch": 0.8925143953934741, + "grad_norm": 31.840341115692503, + "learning_rate": 1.7392836966831553e-08, + "logits/chosen": 0.5662111639976501, + "logits/rejected": 0.5857071876525879, + "logps/chosen": -462.5718688964844, + "logps/rejected": -586.9605712890625, + "loss": 0.4635, + "rewards/accuracies": 0.8999999761581421, + "rewards/chosen": -1.7450441122055054, + "rewards/margins": 1.6221263408660889, + "rewards/rejected": -3.3671703338623047, + "step": 3720 + }, + { + "epoch": 0.8949136276391555, + "grad_norm": 33.21694320716617, + "learning_rate": 1.663366935657373e-08, + "logits/chosen": 0.6892956495285034, + "logits/rejected": 0.7154291868209839, + "logps/chosen": -404.4335021972656, + "logps/rejected": -538.8549194335938, + "loss": 0.5278, + "rewards/accuracies": 0.7250000238418579, + "rewards/chosen": -1.5369670391082764, + "rewards/margins": 1.303447961807251, + "rewards/rejected": -2.8404150009155273, + "step": 3730 + }, + { + "epoch": 0.8973128598848369, + "grad_norm": 51.74258645160025, + "learning_rate": 1.5890871467258898e-08, + "logits/chosen": 0.6136958003044128, + "logits/rejected": 0.637305498123169, + "logps/chosen": -522.9058837890625, + "logps/rejected": -606.9166259765625, + "loss": 0.4937, + "rewards/accuracies": 0.625, + "rewards/chosen": -1.7736104726791382, + "rewards/margins": 1.2294580936431885, + "rewards/rejected": -3.003068447113037, + "step": 3740 + }, + { + "epoch": 0.8997120921305183, + "grad_norm": 31.939433243371568, + "learning_rate": 1.5164495403207967e-08, + "logits/chosen": 0.6770081520080566, + "logits/rejected": 0.6946592926979065, + "logps/chosen": -465.00634765625, + "logps/rejected": -675.024169921875, + "loss": 0.4626, + "rewards/accuracies": 0.8999999761581421, + "rewards/chosen": -1.8869049549102783, + "rewards/margins": 1.8580411672592163, + "rewards/rejected": -3.744946002960205, + "step": 3750 + }, + { + "epoch": 0.9021113243761996, + "grad_norm": 32.427853119545475, + "learning_rate": 1.4454592116815962e-08, + "logits/chosen": 0.5860388875007629, + "logits/rejected": 0.6506261825561523, + "logps/chosen": -435.30615234375, + "logps/rejected": -577.177490234375, + "loss": 0.4524, + "rewards/accuracies": 0.75, + "rewards/chosen": -1.5946733951568604, + "rewards/margins": 1.3281992673873901, + "rewards/rejected": -2.922873020172119, + "step": 3760 + }, + { + "epoch": 0.904510556621881, + "grad_norm": 27.60653570560778, + "learning_rate": 1.3761211404977934e-08, + "logits/chosen": 0.6711539030075073, + "logits/rejected": 0.7092609405517578, + "logps/chosen": -433.2408142089844, + "logps/rejected": -622.9896240234375, + "loss": 0.4364, + "rewards/accuracies": 0.75, + "rewards/chosen": -1.802742600440979, + "rewards/margins": 2.034799814224243, + "rewards/rejected": -3.8375420570373535, + "step": 3770 + }, + { + "epoch": 0.9069097888675623, + "grad_norm": 36.488044765002755, + "learning_rate": 1.3084401905596177e-08, + "logits/chosen": 0.7667958736419678, + "logits/rejected": 0.7162773013114929, + "logps/chosen": -481.6435546875, + "logps/rejected": -564.5675659179688, + "loss": 0.4803, + "rewards/accuracies": 0.8500000238418579, + "rewards/chosen": -1.5850133895874023, + "rewards/margins": 1.4455559253692627, + "rewards/rejected": -3.030569553375244, + "step": 3780 + }, + { + "epoch": 0.9093090211132437, + "grad_norm": 46.89570443675896, + "learning_rate": 1.2424211094168053e-08, + "logits/chosen": 0.6919952630996704, + "logits/rejected": 0.6967671513557434, + "logps/chosen": -491.3282165527344, + "logps/rejected": -630.562255859375, + "loss": 0.5139, + "rewards/accuracies": 0.7749999761581421, + "rewards/chosen": -1.4634296894073486, + "rewards/margins": 1.5717954635620117, + "rewards/rejected": -3.0352253913879395, + "step": 3790 + }, + { + "epoch": 0.9117082533589251, + "grad_norm": 38.92483757033139, + "learning_rate": 1.1780685280456143e-08, + "logits/chosen": 0.6587939858436584, + "logits/rejected": 0.6271030902862549, + "logps/chosen": -507.9803161621094, + "logps/rejected": -711.3810424804688, + "loss": 0.5085, + "rewards/accuracies": 0.8500000238418579, + "rewards/chosen": -1.9189660549163818, + "rewards/margins": 2.0694832801818848, + "rewards/rejected": -3.988449811935425, + "step": 3800 + }, + { + "epoch": 0.9141074856046065, + "grad_norm": 39.99005445096094, + "learning_rate": 1.1153869605239564e-08, + "logits/chosen": 0.6913423538208008, + "logits/rejected": 0.7273651957511902, + "logps/chosen": -432.20458984375, + "logps/rejected": -463.5044860839844, + "loss": 0.4905, + "rewards/accuracies": 0.7250000238418579, + "rewards/chosen": -1.4659842252731323, + "rewards/margins": 0.7735913991928101, + "rewards/rejected": -2.2395756244659424, + "step": 3810 + }, + { + "epoch": 0.9165067178502879, + "grad_norm": 42.80930835029662, + "learning_rate": 1.0543808037147606e-08, + "logits/chosen": 0.6128347516059875, + "logits/rejected": 0.7182141542434692, + "logps/chosen": -448.8204040527344, + "logps/rejected": -642.3395385742188, + "loss": 0.4698, + "rewards/accuracies": 0.8500000238418579, + "rewards/chosen": -1.679520606994629, + "rewards/margins": 1.9481945037841797, + "rewards/rejected": -3.6277146339416504, + "step": 3820 + }, + { + "epoch": 0.9189059500959693, + "grad_norm": 29.4673000753317, + "learning_rate": 9.95054336957557e-09, + "logits/chosen": 0.5145514011383057, + "logits/rejected": 0.604987621307373, + "logps/chosen": -407.84942626953125, + "logps/rejected": -521.2293701171875, + "loss": 0.4726, + "rewards/accuracies": 0.800000011920929, + "rewards/chosen": -1.442375898361206, + "rewards/margins": 1.062880277633667, + "rewards/rejected": -2.505256175994873, + "step": 3830 + }, + { + "epoch": 0.9213051823416507, + "grad_norm": 47.79956261456909, + "learning_rate": 9.37411721768286e-09, + "logits/chosen": 0.7111871242523193, + "logits/rejected": 0.7023571729660034, + "logps/chosen": -472.0362243652344, + "logps/rejected": -679.4020385742188, + "loss": 0.4726, + "rewards/accuracies": 0.824999988079071, + "rewards/chosen": -1.7883527278900146, + "rewards/margins": 1.815100073814392, + "rewards/rejected": -3.6034531593322754, + "step": 3840 + }, + { + "epoch": 0.9237044145873321, + "grad_norm": 35.38700307758538, + "learning_rate": 8.81457001547392e-09, + "logits/chosen": 0.5503877997398376, + "logits/rejected": 0.5560935735702515, + "logps/chosen": -431.03704833984375, + "logps/rejected": -531.4757690429688, + "loss": 0.4753, + "rewards/accuracies": 0.824999988079071, + "rewards/chosen": -1.6167274713516235, + "rewards/margins": 0.931064248085022, + "rewards/rejected": -2.5477917194366455, + "step": 3850 + }, + { + "epoch": 0.9261036468330134, + "grad_norm": 34.31827386168351, + "learning_rate": 8.271941012961942e-09, + "logits/chosen": 0.6740199327468872, + "logits/rejected": 0.696508526802063, + "logps/chosen": -416.02911376953125, + "logps/rejected": -684.9498901367188, + "loss": 0.4683, + "rewards/accuracies": 0.75, + "rewards/chosen": -1.8001444339752197, + "rewards/margins": 2.1946005821228027, + "rewards/rejected": -3.9947447776794434, + "step": 3860 + }, + { + "epoch": 0.9285028790786948, + "grad_norm": 31.376350645526617, + "learning_rate": 7.746268273415568e-09, + "logits/chosen": 0.6382318735122681, + "logits/rejected": 0.6174970865249634, + "logps/chosen": -444.5672302246094, + "logps/rejected": -539.0169677734375, + "loss": 0.485, + "rewards/accuracies": 0.6499999761581421, + "rewards/chosen": -1.6137769222259521, + "rewards/margins": 0.47504907846450806, + "rewards/rejected": -2.0888259410858154, + "step": 3870 + }, + { + "epoch": 0.9309021113243762, + "grad_norm": 30.055393249663297, + "learning_rate": 7.237588670689076e-09, + "logits/chosen": 0.582071840763092, + "logits/rejected": 0.585811972618103, + "logps/chosen": -448.01507568359375, + "logps/rejected": -605.9859008789062, + "loss": 0.4628, + "rewards/accuracies": 0.7250000238418579, + "rewards/chosen": -1.6493518352508545, + "rewards/margins": 2.0434372425079346, + "rewards/rejected": -3.692789077758789, + "step": 3880 + }, + { + "epoch": 0.9333013435700576, + "grad_norm": 37.008005737101136, + "learning_rate": 6.745937886635606e-09, + "logits/chosen": 0.5251346230506897, + "logits/rejected": 0.5287281274795532, + "logps/chosen": -482.1429138183594, + "logps/rejected": -692.1409301757812, + "loss": 0.4634, + "rewards/accuracies": 0.7749999761581421, + "rewards/chosen": -1.7261474132537842, + "rewards/margins": 2.155860424041748, + "rewards/rejected": -3.8820080757141113, + "step": 3890 + }, + { + "epoch": 0.935700575815739, + "grad_norm": 42.882926576601946, + "learning_rate": 6.271350408604409e-09, + "logits/chosen": 0.5571666359901428, + "logits/rejected": 0.5792709589004517, + "logps/chosen": -371.2899475097656, + "logps/rejected": -534.5432739257812, + "loss": 0.5064, + "rewards/accuracies": 0.8999999761581421, + "rewards/chosen": -1.3483814001083374, + "rewards/margins": 1.4740469455718994, + "rewards/rejected": -2.8224284648895264, + "step": 3900 + }, + { + "epoch": 0.9380998080614203, + "grad_norm": 31.83370209912519, + "learning_rate": 5.813859527021487e-09, + "logits/chosen": 0.7409833073616028, + "logits/rejected": 0.7550204396247864, + "logps/chosen": -449.55535888671875, + "logps/rejected": -595.6565551757812, + "loss": 0.4513, + "rewards/accuracies": 0.8999999761581421, + "rewards/chosen": -1.6462414264678955, + "rewards/margins": 1.765639066696167, + "rewards/rejected": -3.4118804931640625, + "step": 3910 + }, + { + "epoch": 0.9404990403071017, + "grad_norm": 29.90782487751284, + "learning_rate": 5.373497333054616e-09, + "logits/chosen": 0.6009663343429565, + "logits/rejected": 0.6457256078720093, + "logps/chosen": -460.5867614746094, + "logps/rejected": -528.3399047851562, + "loss": 0.5268, + "rewards/accuracies": 0.75, + "rewards/chosen": -1.7170873880386353, + "rewards/margins": 0.7792832851409912, + "rewards/rejected": -2.496370792388916, + "step": 3920 + }, + { + "epoch": 0.9428982725527831, + "grad_norm": 30.797799497811695, + "learning_rate": 4.950294716362213e-09, + "logits/chosen": 0.7179510593414307, + "logits/rejected": 0.7142021059989929, + "logps/chosen": -480.66156005859375, + "logps/rejected": -584.0746459960938, + "loss": 0.4984, + "rewards/accuracies": 0.699999988079071, + "rewards/chosen": -1.6999647617340088, + "rewards/margins": 1.0811774730682373, + "rewards/rejected": -2.781142234802246, + "step": 3930 + }, + { + "epoch": 0.9452975047984645, + "grad_norm": 32.61397088226819, + "learning_rate": 4.544281362926422e-09, + "logits/chosen": 0.5735732316970825, + "logits/rejected": 0.624186635017395, + "logps/chosen": -483.68231201171875, + "logps/rejected": -595.20166015625, + "loss": 0.5372, + "rewards/accuracies": 0.7749999761581421, + "rewards/chosen": -1.586411714553833, + "rewards/margins": 1.3250799179077148, + "rewards/rejected": -2.911491870880127, + "step": 3940 + }, + { + "epoch": 0.9476967370441459, + "grad_norm": 31.88712634446775, + "learning_rate": 4.15548575297095e-09, + "logits/chosen": 0.695099949836731, + "logits/rejected": 0.7135946154594421, + "logps/chosen": -465.44671630859375, + "logps/rejected": -647.7462768554688, + "loss": 0.4531, + "rewards/accuracies": 0.7749999761581421, + "rewards/chosen": -1.9738279581069946, + "rewards/margins": 1.9751651287078857, + "rewards/rejected": -3.94899320602417, + "step": 3950 + }, + { + "epoch": 0.9500959692898272, + "grad_norm": 26.750814685545, + "learning_rate": 3.7839351589631366e-09, + "logits/chosen": 0.5124061703681946, + "logits/rejected": 0.5727663040161133, + "logps/chosen": -414.16766357421875, + "logps/rejected": -561.9720458984375, + "loss": 0.4955, + "rewards/accuracies": 0.7749999761581421, + "rewards/chosen": -1.8436729907989502, + "rewards/margins": 0.9941961169242859, + "rewards/rejected": -2.837869167327881, + "step": 3960 + }, + { + "epoch": 0.9524952015355086, + "grad_norm": 33.88554208031626, + "learning_rate": 3.4296556437010405e-09, + "logits/chosen": 0.6761840581893921, + "logits/rejected": 0.7198007702827454, + "logps/chosen": -418.03204345703125, + "logps/rejected": -517.2802734375, + "loss": 0.4897, + "rewards/accuracies": 0.7749999761581421, + "rewards/chosen": -1.9059219360351562, + "rewards/margins": 1.1033378839492798, + "rewards/rejected": -3.0092597007751465, + "step": 3970 + }, + { + "epoch": 0.95489443378119, + "grad_norm": 35.312495401161826, + "learning_rate": 3.092672058485124e-09, + "logits/chosen": 0.60147625207901, + "logits/rejected": 0.6507999300956726, + "logps/chosen": -451.183349609375, + "logps/rejected": -656.1090087890625, + "loss": 0.5143, + "rewards/accuracies": 0.800000011920929, + "rewards/chosen": -2.024777889251709, + "rewards/margins": 1.9228811264038086, + "rewards/rejected": -3.9476592540740967, + "step": 3980 + }, + { + "epoch": 0.9572936660268714, + "grad_norm": 40.30857026059499, + "learning_rate": 2.7730080413750356e-09, + "logits/chosen": 0.6580095887184143, + "logits/rejected": 0.61357581615448, + "logps/chosen": -445.29559326171875, + "logps/rejected": -532.5142211914062, + "loss": 0.5021, + "rewards/accuracies": 0.7749999761581421, + "rewards/chosen": -1.4763895273208618, + "rewards/margins": 0.9797104001045227, + "rewards/rejected": -2.4560999870300293, + "step": 3990 + }, + { + "epoch": 0.9596928982725528, + "grad_norm": 29.89336340827424, + "learning_rate": 2.4706860155316033e-09, + "logits/chosen": 0.6960023641586304, + "logits/rejected": 0.6264194250106812, + "logps/chosen": -556.0472412109375, + "logps/rejected": -656.1188354492188, + "loss": 0.4991, + "rewards/accuracies": 0.699999988079071, + "rewards/chosen": -1.868370771408081, + "rewards/margins": 1.122097134590149, + "rewards/rejected": -2.9904677867889404, + "step": 4000 + }, + { + "epoch": 0.9596928982725528, + "eval_logits/chosen": 0.7145485281944275, + "eval_logits/rejected": 0.7296622395515442, + "eval_logps/chosen": -449.66363525390625, + "eval_logps/rejected": -602.4407958984375, + "eval_loss": 0.48357322812080383, + "eval_rewards/accuracies": 0.7892857193946838, + "eval_rewards/chosen": -1.707390308380127, + "eval_rewards/margins": 1.523337721824646, + "eval_rewards/rejected": -3.2307281494140625, + "eval_runtime": 86.6873, + "eval_samples_per_second": 51.461, + "eval_steps_per_second": 0.808, + "step": 4000 + }, + { + "epoch": 0.9620921305182342, + "grad_norm": 58.07953756761719, + "learning_rate": 2.185727187643843e-09, + "logits/chosen": 0.6966633200645447, + "logits/rejected": 0.7185578942298889, + "logps/chosen": -392.63397216796875, + "logps/rejected": -583.4874267578125, + "loss": 0.4796, + "rewards/accuracies": 0.949999988079071, + "rewards/chosen": -1.5413339138031006, + "rewards/margins": 1.9730831384658813, + "rewards/rejected": -3.5144171714782715, + "step": 4010 + }, + { + "epoch": 0.9644913627639156, + "grad_norm": 46.10404234203207, + "learning_rate": 1.9181515464413434e-09, + "logits/chosen": 0.6493710279464722, + "logits/rejected": 0.6657570004463196, + "logps/chosen": -575.0656127929688, + "logps/rejected": -729.4915771484375, + "loss": 0.5193, + "rewards/accuracies": 0.800000011920929, + "rewards/chosen": -2.0319643020629883, + "rewards/margins": 1.6315898895263672, + "rewards/rejected": -3.6635544300079346, + "step": 4020 + }, + { + "epoch": 0.966890595009597, + "grad_norm": 39.22948436791097, + "learning_rate": 1.6679778612923302e-09, + "logits/chosen": 0.5851081013679504, + "logits/rejected": 0.5799533128738403, + "logps/chosen": -480.31463623046875, + "logps/rejected": -570.2797241210938, + "loss": 0.4747, + "rewards/accuracies": 0.699999988079071, + "rewards/chosen": -1.7180426120758057, + "rewards/margins": 0.8774716258049011, + "rewards/rejected": -2.5955138206481934, + "step": 4030 + }, + { + "epoch": 0.9692898272552783, + "grad_norm": 50.62146798404109, + "learning_rate": 1.43522368088686e-09, + "logits/chosen": 0.5723304748535156, + "logits/rejected": 0.6564613580703735, + "logps/chosen": -469.9656677246094, + "logps/rejected": -674.6346435546875, + "loss": 0.5392, + "rewards/accuracies": 0.824999988079071, + "rewards/chosen": -1.904486894607544, + "rewards/margins": 2.1075260639190674, + "rewards/rejected": -4.012012481689453, + "step": 4040 + }, + { + "epoch": 0.9716890595009597, + "grad_norm": 49.07423025714521, + "learning_rate": 1.2199053320059993e-09, + "logits/chosen": 0.6182990074157715, + "logits/rejected": 0.6073701977729797, + "logps/chosen": -455.4971618652344, + "logps/rejected": -613.2784423828125, + "loss": 0.487, + "rewards/accuracies": 0.824999988079071, + "rewards/chosen": -1.631026029586792, + "rewards/margins": 1.453333854675293, + "rewards/rejected": -3.084359645843506, + "step": 4050 + }, + { + "epoch": 0.974088291746641, + "grad_norm": 27.42422616328582, + "learning_rate": 1.0220379183764338e-09, + "logits/chosen": 0.58095383644104, + "logits/rejected": 0.6131027340888977, + "logps/chosen": -382.36871337890625, + "logps/rejected": -575.1259155273438, + "loss": 0.4948, + "rewards/accuracies": 0.7749999761581421, + "rewards/chosen": -1.5139602422714233, + "rewards/margins": 1.8817888498306274, + "rewards/rejected": -3.3957488536834717, + "step": 4060 + }, + { + "epoch": 0.9764875239923224, + "grad_norm": 43.846222208955815, + "learning_rate": 8.416353196111503e-10, + "logits/chosen": 0.7621703147888184, + "logits/rejected": 0.7751132845878601, + "logps/chosen": -456.3731994628906, + "logps/rejected": -598.9076538085938, + "loss": 0.5192, + "rewards/accuracies": 0.699999988079071, + "rewards/chosen": -1.952593207359314, + "rewards/margins": 1.5192127227783203, + "rewards/rejected": -3.471806049346924, + "step": 4070 + }, + { + "epoch": 0.9788867562380038, + "grad_norm": 40.952265372902325, + "learning_rate": 6.787101902356873e-10, + "logits/chosen": 0.6373910307884216, + "logits/rejected": 0.6397606730461121, + "logps/chosen": -465.97174072265625, + "logps/rejected": -629.4970092773438, + "loss": 0.4754, + "rewards/accuracies": 0.7250000238418579, + "rewards/chosen": -1.8233009576797485, + "rewards/margins": 1.4212223291397095, + "rewards/rejected": -3.2445228099823, + "step": 4080 + }, + { + "epoch": 0.9812859884836852, + "grad_norm": 47.62972519366363, + "learning_rate": 5.332739588005953e-10, + "logits/chosen": 0.5775616765022278, + "logits/rejected": 0.603803277015686, + "logps/chosen": -378.0145568847656, + "logps/rejected": -548.2694702148438, + "loss": 0.4918, + "rewards/accuracies": 0.824999988079071, + "rewards/chosen": -1.5502240657806396, + "rewards/margins": 1.536120057106018, + "rewards/rejected": -3.086343765258789, + "step": 4090 + }, + { + "epoch": 0.9836852207293666, + "grad_norm": 43.81658284237901, + "learning_rate": 4.053368270797164e-10, + "logits/chosen": 0.6224797368049622, + "logits/rejected": 0.6658841371536255, + "logps/chosen": -455.75555419921875, + "logps/rejected": -602.9154663085938, + "loss": 0.4775, + "rewards/accuracies": 0.7250000238418579, + "rewards/chosen": -2.0334808826446533, + "rewards/margins": 1.5407496690750122, + "rewards/rejected": -3.574230670928955, + "step": 4100 + }, + { + "epoch": 0.986084452975048, + "grad_norm": 26.2080571697754, + "learning_rate": 2.949077693545354e-10, + "logits/chosen": 0.5944604873657227, + "logits/rejected": 0.6213740110397339, + "logps/chosen": -458.01312255859375, + "logps/rejected": -585.683349609375, + "loss": 0.5377, + "rewards/accuracies": 0.6499999761581421, + "rewards/chosen": -1.6532799005508423, + "rewards/margins": 0.9707270860671997, + "rewards/rejected": -2.624006748199463, + "step": 4110 + }, + { + "epoch": 0.9884836852207294, + "grad_norm": 55.389028996175746, + "learning_rate": 2.0199453178471047e-10, + "logits/chosen": 0.6714688539505005, + "logits/rejected": 0.5972020030021667, + "logps/chosen": -501.0025329589844, + "logps/rejected": -521.9111938476562, + "loss": 0.5007, + "rewards/accuracies": 0.800000011920929, + "rewards/chosen": -1.590149998664856, + "rewards/margins": 0.6970120072364807, + "rewards/rejected": -2.2871620655059814, + "step": 4120 + }, + { + "epoch": 0.9908829174664108, + "grad_norm": 32.72649040496507, + "learning_rate": 1.266036318647301e-10, + "logits/chosen": 0.5367740392684937, + "logits/rejected": 0.5260412693023682, + "logps/chosen": -512.4938354492188, + "logps/rejected": -691.9700927734375, + "loss": 0.4886, + "rewards/accuracies": 0.8999999761581421, + "rewards/chosen": -1.8918201923370361, + "rewards/margins": 2.05039119720459, + "rewards/rejected": -3.942211627960205, + "step": 4130 + }, + { + "epoch": 0.9932821497120922, + "grad_norm": 41.73724251859538, + "learning_rate": 6.874035796672339e-11, + "logits/chosen": 0.7122509479522705, + "logits/rejected": 0.6869832277297974, + "logps/chosen": -482.57952880859375, + "logps/rejected": -656.2966918945312, + "loss": 0.4923, + "rewards/accuracies": 0.824999988079071, + "rewards/chosen": -1.6301801204681396, + "rewards/margins": 2.3234925270080566, + "rewards/rejected": -3.953672409057617, + "step": 4140 + }, + { + "epoch": 0.9956813819577736, + "grad_norm": 40.29970530677951, + "learning_rate": 2.8408768969423458e-11, + "logits/chosen": 0.6537809371948242, + "logits/rejected": 0.6641941070556641, + "logps/chosen": -462.24432373046875, + "logps/rejected": -581.655029296875, + "loss": 0.5032, + "rewards/accuracies": 0.75, + "rewards/chosen": -1.6611082553863525, + "rewards/margins": 1.1628700494766235, + "rewards/rejected": -2.8239779472351074, + "step": 4150 + }, + { + "epoch": 0.9980806142034548, + "grad_norm": 42.46853911862954, + "learning_rate": 5.611693973617271e-12, + "logits/chosen": 0.6924387216567993, + "logits/rejected": 0.6893667578697205, + "logps/chosen": -396.61761474609375, + "logps/rejected": -547.9915771484375, + "loss": 0.5326, + "rewards/accuracies": 0.75, + "rewards/chosen": -1.5040003061294556, + "rewards/margins": 1.3634313344955444, + "rewards/rejected": -2.867431640625, + "step": 4160 + }, + { + "epoch": 1.0, + "step": 4168, + "total_flos": 0.0, + "train_loss": 0.5339851984021302, + "train_runtime": 6933.6865, + "train_samples_per_second": 19.235, + "train_steps_per_second": 0.601 + } + ], + "logging_steps": 10, + "max_steps": 4168, + "num_input_tokens_seen": 0, + "num_train_epochs": 1, + "save_steps": 10000, + "stateful_callbacks": { + "TrainerControl": { + "args": { + "should_epoch_stop": false, + "should_evaluate": false, + "should_log": false, + "should_save": true, + "should_training_stop": true + }, + "attributes": {} + } + }, + "total_flos": 0.0, + "train_batch_size": 4, + "trial_name": null, + "trial_params": null +}