{ "best_metric": null, "best_model_checkpoint": null, "epoch": 1.0, "eval_steps": 100, "global_step": 637, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.0015698587127158557, "grad_norm": 0.982062816619873, "learning_rate": 7.8125e-08, "logits/chosen": -1.9796161651611328, "logits/rejected": -1.7990202903747559, "logps/chosen": -276.0213317871094, "logps/rejected": -220.55433654785156, "loss": 0.6931, "rewards/accuracies": 0.0, "rewards/chosen": 0.0, "rewards/margins": 0.0, "rewards/rejected": 0.0, "step": 1 }, { "epoch": 0.015698587127158554, "grad_norm": 0.9399773478507996, "learning_rate": 7.8125e-07, "logits/chosen": -2.2138850688934326, "logits/rejected": -2.1262121200561523, "logps/chosen": -290.8321533203125, "logps/rejected": -228.361328125, "loss": 0.6929, "rewards/accuracies": 0.4652777910232544, "rewards/chosen": 0.00492230337113142, "rewards/margins": 0.0006595260347239673, "rewards/rejected": 0.0042627775110304356, "step": 10 }, { "epoch": 0.03139717425431711, "grad_norm": 1.019684076309204, "learning_rate": 1.5625e-06, "logits/chosen": -2.3072190284729004, "logits/rejected": -2.201758861541748, "logps/chosen": -265.79144287109375, "logps/rejected": -252.1865997314453, "loss": 0.6895, "rewards/accuracies": 0.643750011920929, "rewards/chosen": 0.029890870675444603, "rewards/margins": 0.009029369801282883, "rewards/rejected": 0.02086150087416172, "step": 20 }, { "epoch": 0.04709576138147567, "grad_norm": 0.9330567121505737, "learning_rate": 2.3437500000000002e-06, "logits/chosen": -2.312687635421753, "logits/rejected": -2.206019878387451, "logps/chosen": -268.0202331542969, "logps/rejected": -250.2025146484375, "loss": 0.6844, "rewards/accuracies": 0.6499999761581421, "rewards/chosen": 0.05118671804666519, "rewards/margins": 0.019565049558877945, "rewards/rejected": 0.03162166848778725, "step": 30 }, { "epoch": 0.06279434850863422, "grad_norm": 0.9986972808837891, "learning_rate": 3.125e-06, "logits/chosen": -2.2849414348602295, "logits/rejected": -2.1605541706085205, "logps/chosen": -248.8426971435547, "logps/rejected": -222.5389404296875, "loss": 0.675, "rewards/accuracies": 0.6812499761581421, "rewards/chosen": 0.06607525795698166, "rewards/margins": 0.039594851434230804, "rewards/rejected": 0.026480406522750854, "step": 40 }, { "epoch": 0.07849293563579278, "grad_norm": 1.18948495388031, "learning_rate": 3.90625e-06, "logits/chosen": -2.237248659133911, "logits/rejected": -2.190307855606079, "logps/chosen": -268.77899169921875, "logps/rejected": -278.4407958984375, "loss": 0.6605, "rewards/accuracies": 0.71875, "rewards/chosen": 0.00511029502376914, "rewards/margins": 0.05497710779309273, "rewards/rejected": -0.04986681044101715, "step": 50 }, { "epoch": 0.09419152276295134, "grad_norm": 1.285398006439209, "learning_rate": 4.6875000000000004e-06, "logits/chosen": -2.265840768814087, "logits/rejected": -2.1778528690338135, "logps/chosen": -272.609130859375, "logps/rejected": -277.2625732421875, "loss": 0.6429, "rewards/accuracies": 0.71875, "rewards/chosen": -0.07459741830825806, "rewards/margins": 0.12982353568077087, "rewards/rejected": -0.20442095398902893, "step": 60 }, { "epoch": 0.10989010989010989, "grad_norm": 1.8551888465881348, "learning_rate": 4.998647417232375e-06, "logits/chosen": -2.1861605644226074, "logits/rejected": -2.0909242630004883, "logps/chosen": -267.07611083984375, "logps/rejected": -259.22210693359375, "loss": 0.6386, "rewards/accuracies": 0.643750011920929, "rewards/chosen": -0.2929137647151947, "rewards/margins": 0.1055971160531044, "rewards/rejected": -0.3985108435153961, "step": 70 }, { "epoch": 0.12558869701726844, "grad_norm": 1.9893733263015747, "learning_rate": 4.990386933279973e-06, "logits/chosen": -2.184288740158081, "logits/rejected": -2.1073215007781982, "logps/chosen": -290.60943603515625, "logps/rejected": -284.17425537109375, "loss": 0.6057, "rewards/accuracies": 0.699999988079071, "rewards/chosen": -0.30957552790641785, "rewards/margins": 0.2696777582168579, "rewards/rejected": -0.5792533159255981, "step": 80 }, { "epoch": 0.141287284144427, "grad_norm": 3.498678684234619, "learning_rate": 4.974642195009681e-06, "logits/chosen": -2.1730189323425293, "logits/rejected": -2.092939853668213, "logps/chosen": -326.27838134765625, "logps/rejected": -310.67767333984375, "loss": 0.6017, "rewards/accuracies": 0.6499999761581421, "rewards/chosen": -0.5229882001876831, "rewards/margins": 0.20935317873954773, "rewards/rejected": -0.7323412895202637, "step": 90 }, { "epoch": 0.15698587127158556, "grad_norm": 2.8239424228668213, "learning_rate": 4.951460519416228e-06, "logits/chosen": -2.254025936126709, "logits/rejected": -2.1698710918426514, "logps/chosen": -328.6602783203125, "logps/rejected": -333.0082092285156, "loss": 0.6099, "rewards/accuracies": 0.6875, "rewards/chosen": -0.4128250479698181, "rewards/margins": 0.2569137513637543, "rewards/rejected": -0.6697388887405396, "step": 100 }, { "epoch": 0.15698587127158556, "eval_logits/chosen": -2.2190940380096436, "eval_logits/rejected": -2.130751371383667, "eval_logps/chosen": -304.2706604003906, "eval_logps/rejected": -307.0733642578125, "eval_loss": 0.59989333152771, "eval_rewards/accuracies": 0.636904776096344, "eval_rewards/chosen": -0.45707327127456665, "eval_rewards/margins": 0.2184356451034546, "eval_rewards/rejected": -0.675508975982666, "eval_runtime": 219.8057, "eval_samples_per_second": 9.099, "eval_steps_per_second": 0.382, "step": 100 }, { "epoch": 0.1726844583987441, "grad_norm": 4.492749214172363, "learning_rate": 4.920911573406925e-06, "logits/chosen": -2.2060790061950684, "logits/rejected": -2.0495288372039795, "logps/chosen": -324.7986755371094, "logps/rejected": -339.40802001953125, "loss": 0.5822, "rewards/accuracies": 0.7250000238418579, "rewards/chosen": -0.5525876879692078, "rewards/margins": 0.44238606095314026, "rewards/rejected": -0.9949736595153809, "step": 110 }, { "epoch": 0.18838304552590268, "grad_norm": 4.563241481781006, "learning_rate": 4.883087164434672e-06, "logits/chosen": -2.0682525634765625, "logits/rejected": -2.0306172370910645, "logps/chosen": -367.96966552734375, "logps/rejected": -406.1026611328125, "loss": 0.5553, "rewards/accuracies": 0.71875, "rewards/chosen": -0.9666538238525391, "rewards/margins": 0.4744674563407898, "rewards/rejected": -1.4411213397979736, "step": 120 }, { "epoch": 0.20408163265306123, "grad_norm": 4.36543607711792, "learning_rate": 4.8381009645929044e-06, "logits/chosen": -1.958156943321228, "logits/rejected": -1.8605877161026, "logps/chosen": -372.92083740234375, "logps/rejected": -398.33209228515625, "loss": 0.5609, "rewards/accuracies": 0.6812499761581421, "rewards/chosen": -0.9711011052131653, "rewards/margins": 0.45527857542037964, "rewards/rejected": -1.426379680633545, "step": 130 }, { "epoch": 0.21978021978021978, "grad_norm": 3.9676990509033203, "learning_rate": 4.786088169001671e-06, "logits/chosen": -1.9477580785751343, "logits/rejected": -1.9076995849609375, "logps/chosen": -364.2967224121094, "logps/rejected": -398.6386413574219, "loss": 0.5805, "rewards/accuracies": 0.6312500238418579, "rewards/chosen": -1.2606209516525269, "rewards/margins": 0.37748971581459045, "rewards/rejected": -1.6381107568740845, "step": 140 }, { "epoch": 0.23547880690737832, "grad_norm": 4.215748310089111, "learning_rate": 4.727205089511466e-06, "logits/chosen": -1.8670552968978882, "logits/rejected": -1.8230712413787842, "logps/chosen": -380.49237060546875, "logps/rejected": -442.67596435546875, "loss": 0.5594, "rewards/accuracies": 0.6499999761581421, "rewards/chosen": -1.3134872913360596, "rewards/margins": 0.47393256425857544, "rewards/rejected": -1.7874199151992798, "step": 150 }, { "epoch": 0.25117739403453687, "grad_norm": 4.537078380584717, "learning_rate": 4.661628684945851e-06, "logits/chosen": -1.8435173034667969, "logits/rejected": -1.7608001232147217, "logps/chosen": -380.8707580566406, "logps/rejected": -421.7408142089844, "loss": 0.5537, "rewards/accuracies": 0.737500011920929, "rewards/chosen": -1.0851117372512817, "rewards/margins": 0.5639781355857849, "rewards/rejected": -1.6490898132324219, "step": 160 }, { "epoch": 0.2668759811616955, "grad_norm": 4.127157211303711, "learning_rate": 4.5895560292946e-06, "logits/chosen": -1.8356807231903076, "logits/rejected": -1.7069345712661743, "logps/chosen": -406.5726623535156, "logps/rejected": -423.89862060546875, "loss": 0.5395, "rewards/accuracies": 0.737500011920929, "rewards/chosen": -1.350484848022461, "rewards/margins": 0.5463389754295349, "rewards/rejected": -1.8968238830566406, "step": 170 }, { "epoch": 0.282574568288854, "grad_norm": 4.295073986053467, "learning_rate": 4.511203719455588e-06, "logits/chosen": -1.805424690246582, "logits/rejected": -1.675389051437378, "logps/chosen": -404.70721435546875, "logps/rejected": -431.9833984375, "loss": 0.539, "rewards/accuracies": 0.699999988079071, "rewards/chosen": -1.289776086807251, "rewards/margins": 0.7749110460281372, "rewards/rejected": -2.0646872520446777, "step": 180 }, { "epoch": 0.29827315541601257, "grad_norm": 4.170295715332031, "learning_rate": 4.426807224305315e-06, "logits/chosen": -1.7862762212753296, "logits/rejected": -1.635716438293457, "logps/chosen": -372.5980529785156, "logps/rejected": -416.7577209472656, "loss": 0.5637, "rewards/accuracies": 0.762499988079071, "rewards/chosen": -1.134557843208313, "rewards/margins": 0.753765344619751, "rewards/rejected": -1.888323187828064, "step": 190 }, { "epoch": 0.3139717425431711, "grad_norm": 4.384995937347412, "learning_rate": 4.336620177054269e-06, "logits/chosen": -1.7506341934204102, "logits/rejected": -1.6495962142944336, "logps/chosen": -347.5675048828125, "logps/rejected": -391.43731689453125, "loss": 0.5468, "rewards/accuracies": 0.78125, "rewards/chosen": -1.009634256362915, "rewards/margins": 0.6290403604507446, "rewards/rejected": -1.6386747360229492, "step": 200 }, { "epoch": 0.3139717425431711, "eval_logits/chosen": -1.607847809791565, "eval_logits/rejected": -1.5156790018081665, "eval_logps/chosen": -379.1949768066406, "eval_logps/rejected": -409.77398681640625, "eval_loss": 0.5353844165802002, "eval_rewards/accuracies": 0.6815476417541504, "eval_rewards/chosen": -1.2063161134719849, "eval_rewards/margins": 0.49619969725608826, "eval_rewards/rejected": -1.702515721321106, "eval_runtime": 219.5785, "eval_samples_per_second": 9.108, "eval_steps_per_second": 0.383, "step": 200 }, { "epoch": 0.32967032967032966, "grad_norm": 4.753127098083496, "learning_rate": 4.240913613013785e-06, "logits/chosen": -1.6041982173919678, "logits/rejected": -1.4586137533187866, "logps/chosen": -403.050537109375, "logps/rejected": -434.6443786621094, "loss": 0.5589, "rewards/accuracies": 0.7437499761581421, "rewards/chosen": -1.2178317308425903, "rewards/margins": 0.7227036952972412, "rewards/rejected": -1.940535306930542, "step": 210 }, { "epoch": 0.3453689167974882, "grad_norm": 4.243059158325195, "learning_rate": 4.139975155065109e-06, "logits/chosen": -1.545710802078247, "logits/rejected": -1.4544004201889038, "logps/chosen": -359.60174560546875, "logps/rejected": -431.0966796875, "loss": 0.5293, "rewards/accuracies": 0.7437499761581421, "rewards/chosen": -1.2607206106185913, "rewards/margins": 0.6980492472648621, "rewards/rejected": -1.9587697982788086, "step": 220 }, { "epoch": 0.36106750392464676, "grad_norm": 4.250452995300293, "learning_rate": 4.034108149278544e-06, "logits/chosen": -1.5040862560272217, "logits/rejected": -1.419395923614502, "logps/chosen": -438.34979248046875, "logps/rejected": -493.54608154296875, "loss": 0.5378, "rewards/accuracies": 0.7562500238418579, "rewards/chosen": -1.6042639017105103, "rewards/margins": 0.6156054735183716, "rewards/rejected": -2.2198691368103027, "step": 230 }, { "epoch": 0.37676609105180536, "grad_norm": 5.608371734619141, "learning_rate": 3.923630753280358e-06, "logits/chosen": -1.5527160167694092, "logits/rejected": -1.4736815690994263, "logps/chosen": -357.8080749511719, "logps/rejected": -439.68218994140625, "loss": 0.4945, "rewards/accuracies": 0.75, "rewards/chosen": -0.9829418063163757, "rewards/margins": 0.868217945098877, "rewards/rejected": -1.8511598110198975, "step": 240 }, { "epoch": 0.3924646781789639, "grad_norm": 4.367254734039307, "learning_rate": 3.80887498010715e-06, "logits/chosen": -1.4042975902557373, "logits/rejected": -1.283562421798706, "logps/chosen": -437.2142639160156, "logps/rejected": -488.03826904296875, "loss": 0.5232, "rewards/accuracies": 0.6937500238418579, "rewards/chosen": -1.8854888677597046, "rewards/margins": 0.7756179571151733, "rewards/rejected": -2.661106824874878, "step": 250 }, { "epoch": 0.40816326530612246, "grad_norm": 3.6756815910339355, "learning_rate": 3.690185700421145e-06, "logits/chosen": -1.534239411354065, "logits/rejected": -1.4822893142700195, "logps/chosen": -338.0312805175781, "logps/rejected": -372.50274658203125, "loss": 0.5643, "rewards/accuracies": 0.699999988079071, "rewards/chosen": -0.9809848666191101, "rewards/margins": 0.43755441904067993, "rewards/rejected": -1.41853928565979, "step": 260 }, { "epoch": 0.423861852433281, "grad_norm": 4.209615707397461, "learning_rate": 3.567919606085004e-06, "logits/chosen": -1.6105585098266602, "logits/rejected": -1.5416157245635986, "logps/chosen": -353.83245849609375, "logps/rejected": -427.657470703125, "loss": 0.5363, "rewards/accuracies": 0.7749999761581421, "rewards/chosen": -0.8973706364631653, "rewards/margins": 0.6859413385391235, "rewards/rejected": -1.5833120346069336, "step": 270 }, { "epoch": 0.43956043956043955, "grad_norm": 5.80835485458374, "learning_rate": 3.442444138210883e-06, "logits/chosen": -1.3900154829025269, "logits/rejected": -1.286752700805664, "logps/chosen": -383.26959228515625, "logps/rejected": -396.59735107421875, "loss": 0.5434, "rewards/accuracies": 0.675000011920929, "rewards/chosen": -1.367221474647522, "rewards/margins": 0.605404257774353, "rewards/rejected": -1.972625732421875, "step": 280 }, { "epoch": 0.4552590266875981, "grad_norm": 5.807376861572266, "learning_rate": 3.314136382905234e-06, "logits/chosen": -1.4290958642959595, "logits/rejected": -1.3117034435272217, "logps/chosen": -405.759765625, "logps/rejected": -466.0645446777344, "loss": 0.5277, "rewards/accuracies": 0.731249988079071, "rewards/chosen": -1.4295778274536133, "rewards/margins": 0.7734361886978149, "rewards/rejected": -2.2030138969421387, "step": 290 }, { "epoch": 0.47095761381475665, "grad_norm": 4.498719215393066, "learning_rate": 3.1833819380279028e-06, "logits/chosen": -1.4070186614990234, "logits/rejected": -1.3326956033706665, "logps/chosen": -398.8482666015625, "logps/rejected": -497.5279235839844, "loss": 0.5195, "rewards/accuracies": 0.737500011920929, "rewards/chosen": -1.5320394039154053, "rewards/margins": 0.9529226422309875, "rewards/rejected": -2.484961986541748, "step": 300 }, { "epoch": 0.47095761381475665, "eval_logits/chosen": -1.452673077583313, "eval_logits/rejected": -1.3522714376449585, "eval_logps/chosen": -418.378173828125, "eval_logps/rejected": -467.829345703125, "eval_loss": 0.5227062702178955, "eval_rewards/accuracies": 0.7083333134651184, "eval_rewards/chosen": -1.5981483459472656, "eval_rewards/margins": 0.684920608997345, "eval_rewards/rejected": -2.2830684185028076, "eval_runtime": 219.6897, "eval_samples_per_second": 9.104, "eval_steps_per_second": 0.382, "step": 300 }, { "epoch": 0.48665620094191525, "grad_norm": 6.346375465393066, "learning_rate": 3.050573754371228e-06, "logits/chosen": -1.4891157150268555, "logits/rejected": -1.425148844718933, "logps/chosen": -399.7137451171875, "logps/rejected": -429.02960205078125, "loss": 0.5355, "rewards/accuracies": 0.71875, "rewards/chosen": -1.2991411685943604, "rewards/margins": 0.6303215026855469, "rewards/rejected": -1.9294626712799072, "step": 310 }, { "epoch": 0.5023547880690737, "grad_norm": 5.018222808837891, "learning_rate": 2.916110954741667e-06, "logits/chosen": -1.4688920974731445, "logits/rejected": -1.295971155166626, "logps/chosen": -429.50250244140625, "logps/rejected": -448.509521484375, "loss": 0.4945, "rewards/accuracies": 0.731249988079071, "rewards/chosen": -1.343754529953003, "rewards/margins": 0.7055485844612122, "rewards/rejected": -2.0493030548095703, "step": 320 }, { "epoch": 0.5180533751962323, "grad_norm": 5.533540725708008, "learning_rate": 2.7803976344929497e-06, "logits/chosen": -1.3964585065841675, "logits/rejected": -1.2397369146347046, "logps/chosen": -450.6763610839844, "logps/rejected": -513.90234375, "loss": 0.5211, "rewards/accuracies": 0.731249988079071, "rewards/chosen": -1.7644832134246826, "rewards/margins": 0.8988251686096191, "rewards/rejected": -2.663308620452881, "step": 330 }, { "epoch": 0.533751962323391, "grad_norm": 4.254801273345947, "learning_rate": 2.6438416471154277e-06, "logits/chosen": -1.420155644416809, "logits/rejected": -1.355759859085083, "logps/chosen": -393.2152099609375, "logps/rejected": -455.6507263183594, "loss": 0.5032, "rewards/accuracies": 0.731249988079071, "rewards/chosen": -1.4955414533615112, "rewards/margins": 0.7609378099441528, "rewards/rejected": -2.256479501724243, "step": 340 }, { "epoch": 0.5494505494505495, "grad_norm": 5.497036933898926, "learning_rate": 2.5068533785312673e-06, "logits/chosen": -1.2714860439300537, "logits/rejected": -1.206830382347107, "logps/chosen": -424.79864501953125, "logps/rejected": -484.2491760253906, "loss": 0.5265, "rewards/accuracies": 0.731249988079071, "rewards/chosen": -1.5535635948181152, "rewards/margins": 0.7852663993835449, "rewards/rejected": -2.33882999420166, "step": 350 }, { "epoch": 0.565149136577708, "grad_norm": 4.5596208572387695, "learning_rate": 2.369844513779026e-06, "logits/chosen": -1.2018563747406006, "logits/rejected": -1.115635871887207, "logps/chosen": -440.96942138671875, "logps/rejected": -488.6031799316406, "loss": 0.5316, "rewards/accuracies": 0.762499988079071, "rewards/chosen": -1.7989181280136108, "rewards/margins": 0.8467502593994141, "rewards/rejected": -2.6456682682037354, "step": 360 }, { "epoch": 0.5808477237048666, "grad_norm": 4.705105781555176, "learning_rate": 2.2332267997940514e-06, "logits/chosen": -1.3089182376861572, "logits/rejected": -1.1716320514678955, "logps/chosen": -400.7124938964844, "logps/rejected": -441.3446350097656, "loss": 0.5463, "rewards/accuracies": 0.762499988079071, "rewards/chosen": -1.4626152515411377, "rewards/margins": 0.795612096786499, "rewards/rejected": -2.2582273483276367, "step": 370 }, { "epoch": 0.5965463108320251, "grad_norm": 4.903963565826416, "learning_rate": 2.097410808002869e-06, "logits/chosen": -1.3345696926116943, "logits/rejected": -1.196610927581787, "logps/chosen": -416.953125, "logps/rejected": -442.70611572265625, "loss": 0.5065, "rewards/accuracies": 0.762499988079071, "rewards/chosen": -1.265023946762085, "rewards/margins": 0.7792559862136841, "rewards/rejected": -2.0442800521850586, "step": 380 }, { "epoch": 0.6122448979591837, "grad_norm": 5.188351631164551, "learning_rate": 1.962804700450265e-06, "logits/chosen": -1.2588812112808228, "logits/rejected": -1.1723003387451172, "logps/chosen": -375.08428955078125, "logps/rejected": -457.165771484375, "loss": 0.5228, "rewards/accuracies": 0.8187500238418579, "rewards/chosen": -1.3058245182037354, "rewards/margins": 0.9940802454948425, "rewards/rejected": -2.2999048233032227, "step": 390 }, { "epoch": 0.6279434850863422, "grad_norm": 6.60071325302124, "learning_rate": 1.8298130031671974e-06, "logits/chosen": -1.3610254526138306, "logits/rejected": -1.1776959896087646, "logps/chosen": -445.6390686035156, "logps/rejected": -512.9346923828125, "loss": 0.4895, "rewards/accuracies": 0.7749999761581421, "rewards/chosen": -1.6864149570465088, "rewards/margins": 0.9683998823165894, "rewards/rejected": -2.6548147201538086, "step": 400 }, { "epoch": 0.6279434850863422, "eval_logits/chosen": -1.2180150747299194, "eval_logits/rejected": -1.1070232391357422, "eval_logps/chosen": -444.1170959472656, "eval_logps/rejected": -506.0622253417969, "eval_loss": 0.514235258102417, "eval_rewards/accuracies": 0.699404776096344, "eval_rewards/chosen": -1.8555371761322021, "eval_rewards/margins": 0.809860348701477, "eval_rewards/rejected": -2.6653974056243896, "eval_runtime": 220.0748, "eval_samples_per_second": 9.088, "eval_steps_per_second": 0.382, "step": 400 }, { "epoch": 0.6436420722135008, "grad_norm": 3.98221755027771, "learning_rate": 1.6988353904658495e-06, "logits/chosen": -1.2132154703140259, "logits/rejected": -1.1685426235198975, "logps/chosen": -435.8744201660156, "logps/rejected": -512.9607543945312, "loss": 0.5104, "rewards/accuracies": 0.7437499761581421, "rewards/chosen": -1.587599515914917, "rewards/margins": 1.0233564376831055, "rewards/rejected": -2.6109557151794434, "step": 410 }, { "epoch": 0.6593406593406593, "grad_norm": 6.315725803375244, "learning_rate": 1.5702654838153641e-06, "logits/chosen": -1.1728157997131348, "logits/rejected": -1.1026521921157837, "logps/chosen": -421.05963134765625, "logps/rejected": -512.4940185546875, "loss": 0.4971, "rewards/accuracies": 0.731249988079071, "rewards/chosen": -1.7201759815216064, "rewards/margins": 0.8986526727676392, "rewards/rejected": -2.618828535079956, "step": 420 }, { "epoch": 0.6750392464678179, "grad_norm": 5.234571933746338, "learning_rate": 1.4444896689079142e-06, "logits/chosen": -1.1644630432128906, "logits/rejected": -1.0723572969436646, "logps/chosen": -465.34661865234375, "logps/rejected": -545.7144165039062, "loss": 0.5029, "rewards/accuracies": 0.768750011920929, "rewards/chosen": -1.9419937133789062, "rewards/margins": 1.0666325092315674, "rewards/rejected": -3.0086259841918945, "step": 430 }, { "epoch": 0.6907378335949764, "grad_norm": 5.252328395843506, "learning_rate": 1.3218859344701634e-06, "logits/chosen": -1.1691018342971802, "logits/rejected": -1.120827317237854, "logps/chosen": -424.86639404296875, "logps/rejected": -522.9156494140625, "loss": 0.4923, "rewards/accuracies": 0.7437499761581421, "rewards/chosen": -1.887139081954956, "rewards/margins": 1.0102875232696533, "rewards/rejected": -2.8974266052246094, "step": 440 }, { "epoch": 0.706436420722135, "grad_norm": 7.3930206298828125, "learning_rate": 1.2028227363097583e-06, "logits/chosen": -1.1967122554779053, "logits/rejected": -1.157003402709961, "logps/chosen": -483.7662048339844, "logps/rejected": -568.5826416015625, "loss": 0.4961, "rewards/accuracies": 0.7437499761581421, "rewards/chosen": -2.155179023742676, "rewards/margins": 0.9878844022750854, "rewards/rejected": -3.14306378364563, "step": 450 }, { "epoch": 0.7221350078492935, "grad_norm": 3.9247488975524902, "learning_rate": 1.0876578900107053e-06, "logits/chosen": -1.2375166416168213, "logits/rejected": -1.1339408159255981, "logps/chosen": -465.45458984375, "logps/rejected": -551.1287231445312, "loss": 0.4842, "rewards/accuracies": 0.7124999761581421, "rewards/chosen": -1.9650770425796509, "rewards/margins": 1.0327082872390747, "rewards/rejected": -2.9977853298187256, "step": 460 }, { "epoch": 0.7378335949764521, "grad_norm": 7.045323848724365, "learning_rate": 9.767374956053584e-07, "logits/chosen": -1.1633169651031494, "logits/rejected": -1.0927058458328247, "logps/chosen": -469.948486328125, "logps/rejected": -547.9520874023438, "loss": 0.4992, "rewards/accuracies": 0.7250000238418579, "rewards/chosen": -2.16100811958313, "rewards/margins": 0.9476286172866821, "rewards/rejected": -3.1086366176605225, "step": 470 }, { "epoch": 0.7535321821036107, "grad_norm": 5.355714797973633, "learning_rate": 8.703948974546592e-07, "logits/chosen": -1.2294596433639526, "logits/rejected": -1.1127874851226807, "logps/chosen": -500.1397399902344, "logps/rejected": -558.7158813476562, "loss": 0.508, "rewards/accuracies": 0.737500011920929, "rewards/chosen": -2.36497163772583, "rewards/margins": 0.9023281335830688, "rewards/rejected": -3.2672996520996094, "step": 480 }, { "epoch": 0.7692307692307693, "grad_norm": 5.423379421234131, "learning_rate": 7.689496824624526e-07, "logits/chosen": -1.206969976425171, "logits/rejected": -1.0124098062515259, "logps/chosen": -503.8736267089844, "logps/rejected": -560.7037353515625, "loss": 0.5118, "rewards/accuracies": 0.8062499761581421, "rewards/chosen": -2.366530656814575, "rewards/margins": 1.0277622938156128, "rewards/rejected": -3.3942928314208984, "step": 490 }, { "epoch": 0.7849293563579278, "grad_norm": 4.5977630615234375, "learning_rate": 6.7270671963451e-07, "logits/chosen": -1.24196195602417, "logits/rejected": -1.166027307510376, "logps/chosen": -503.1126403808594, "logps/rejected": -574.1822509765625, "loss": 0.4992, "rewards/accuracies": 0.75, "rewards/chosen": -2.4480373859405518, "rewards/margins": 0.874676525592804, "rewards/rejected": -3.322713851928711, "step": 500 }, { "epoch": 0.7849293563579278, "eval_logits/chosen": -1.2588800191879272, "eval_logits/rejected": -1.1520029306411743, "eval_logps/chosen": -491.8629455566406, "eval_logps/rejected": -549.8136596679688, "eval_loss": 0.5018974542617798, "eval_rewards/accuracies": 0.7053571343421936, "eval_rewards/chosen": -2.332996129989624, "eval_rewards/margins": 0.7699158191680908, "eval_rewards/rejected": -3.102912187576294, "eval_runtime": 219.9359, "eval_samples_per_second": 9.094, "eval_steps_per_second": 0.382, "step": 500 }, { "epoch": 0.8006279434850864, "grad_norm": 4.936287879943848, "learning_rate": 5.819552438686238e-07, "logits/chosen": -1.2254358530044556, "logits/rejected": -1.1503589153289795, "logps/chosen": -488.61053466796875, "logps/rejected": -558.0599975585938, "loss": 0.4946, "rewards/accuracies": 0.762499988079071, "rewards/chosen": -2.1799943447113037, "rewards/margins": 0.8831037282943726, "rewards/rejected": -3.063098192214966, "step": 510 }, { "epoch": 0.8163265306122449, "grad_norm": 4.393782615661621, "learning_rate": 4.969679867292276e-07, "logits/chosen": -1.4334200620651245, "logits/rejected": -1.2100530862808228, "logps/chosen": -525.8604736328125, "logps/rejected": -564.2894897460938, "loss": 0.5004, "rewards/accuracies": 0.7250000238418579, "rewards/chosen": -2.246795177459717, "rewards/margins": 0.9003856778144836, "rewards/rejected": -3.1471807956695557, "step": 520 }, { "epoch": 0.8320251177394035, "grad_norm": 4.474452495574951, "learning_rate": 4.1800035681877765e-07, "logits/chosen": -1.165138840675354, "logits/rejected": -1.085632085800171, "logps/chosen": -455.230712890625, "logps/rejected": -550.22412109375, "loss": 0.5097, "rewards/accuracies": 0.737500011920929, "rewards/chosen": -2.193859338760376, "rewards/margins": 0.9264294505119324, "rewards/rejected": -3.120288372039795, "step": 530 }, { "epoch": 0.847723704866562, "grad_norm": 4.0765767097473145, "learning_rate": 3.4528967220911287e-07, "logits/chosen": -1.276033639907837, "logits/rejected": -1.2467362880706787, "logps/chosen": -488.040283203125, "logps/rejected": -554.4703979492188, "loss": 0.5123, "rewards/accuracies": 0.706250011920929, "rewards/chosen": -2.10200834274292, "rewards/margins": 0.8350952863693237, "rewards/rejected": -2.937103748321533, "step": 540 }, { "epoch": 0.8634222919937206, "grad_norm": 4.756001949310303, "learning_rate": 2.7905444723949765e-07, "logits/chosen": -1.2069323062896729, "logits/rejected": -1.192297101020813, "logps/chosen": -484.1224060058594, "logps/rejected": -570.1021118164062, "loss": 0.5144, "rewards/accuracies": 0.7749999761581421, "rewards/chosen": -2.127053737640381, "rewards/margins": 0.9687838554382324, "rewards/rejected": -3.095837354660034, "step": 550 }, { "epoch": 0.8791208791208791, "grad_norm": 4.719667911529541, "learning_rate": 2.1949373582475065e-07, "logits/chosen": -1.2955992221832275, "logits/rejected": -1.2129268646240234, "logps/chosen": -453.64794921875, "logps/rejected": -541.0159912109375, "loss": 0.4979, "rewards/accuracies": 0.793749988079071, "rewards/chosen": -1.9799339771270752, "rewards/margins": 1.0608704090118408, "rewards/rejected": -3.040804386138916, "step": 560 }, { "epoch": 0.8948194662480377, "grad_norm": 5.24905252456665, "learning_rate": 1.667865332469379e-07, "logits/chosen": -1.3544114828109741, "logits/rejected": -1.2046071290969849, "logps/chosen": -456.50482177734375, "logps/rejected": -518.5260620117188, "loss": 0.4768, "rewards/accuracies": 0.84375, "rewards/chosen": -1.8672730922698975, "rewards/margins": 1.0792367458343506, "rewards/rejected": -2.946509838104248, "step": 570 }, { "epoch": 0.9105180533751962, "grad_norm": 5.04511833190918, "learning_rate": 1.2109123822844653e-07, "logits/chosen": -1.314582109451294, "logits/rejected": -1.1942718029022217, "logps/chosen": -491.97967529296875, "logps/rejected": -532.6231689453125, "loss": 0.4948, "rewards/accuracies": 0.71875, "rewards/chosen": -2.1609413623809814, "rewards/margins": 0.8695161938667297, "rewards/rejected": -3.0304577350616455, "step": 580 }, { "epoch": 0.9262166405023547, "grad_norm": 4.632051467895508, "learning_rate": 8.254517690300946e-08, "logits/chosen": -1.3019325733184814, "logits/rejected": -1.1224420070648193, "logps/chosen": -463.02874755859375, "logps/rejected": -560.708251953125, "loss": 0.4826, "rewards/accuracies": 0.737500011920929, "rewards/chosen": -2.165086507797241, "rewards/margins": 1.022458791732788, "rewards/rejected": -3.1875455379486084, "step": 590 }, { "epoch": 0.9419152276295133, "grad_norm": 5.119118690490723, "learning_rate": 5.126419011529993e-08, "logits/chosen": -1.251976490020752, "logits/rejected": -1.1646479368209839, "logps/chosen": -498.5111389160156, "logps/rejected": -600.1183471679688, "loss": 0.5001, "rewards/accuracies": 0.737500011920929, "rewards/chosen": -2.133627414703369, "rewards/margins": 1.0616191625595093, "rewards/rejected": -3.195246696472168, "step": 600 }, { "epoch": 0.9419152276295133, "eval_logits/chosen": -1.2825318574905396, "eval_logits/rejected": -1.1767752170562744, "eval_logps/chosen": -475.6836853027344, "eval_logps/rejected": -535.6056518554688, "eval_loss": 0.5020818114280701, "eval_rewards/accuracies": 0.7083333134651184, "eval_rewards/chosen": -2.171203136444092, "eval_rewards/margins": 0.7896292805671692, "eval_rewards/rejected": -2.960832118988037, "eval_runtime": 219.3018, "eval_samples_per_second": 9.12, "eval_steps_per_second": 0.383, "step": 600 }, { "epoch": 0.957613814756672, "grad_norm": 5.81568717956543, "learning_rate": 2.734228528934679e-08, "logits/chosen": -1.3186975717544556, "logits/rejected": -1.1690412759780884, "logps/chosen": -448.78814697265625, "logps/rejected": -523.2693481445312, "loss": 0.5127, "rewards/accuracies": 0.800000011920929, "rewards/chosen": -2.0488333702087402, "rewards/margins": 1.0742355585098267, "rewards/rejected": -3.1230690479278564, "step": 610 }, { "epoch": 0.9733124018838305, "grad_norm": 5.066900253295898, "learning_rate": 1.0851353912008644e-08, "logits/chosen": -1.1879260540008545, "logits/rejected": -1.057185411453247, "logps/chosen": -493.63323974609375, "logps/rejected": -551.7728271484375, "loss": 0.4958, "rewards/accuracies": 0.706250011920929, "rewards/chosen": -2.1862809658050537, "rewards/margins": 0.9377900958061218, "rewards/rejected": -3.1240711212158203, "step": 620 }, { "epoch": 0.989010989010989, "grad_norm": 5.623207092285156, "learning_rate": 1.8409554805329243e-09, "logits/chosen": -1.2257732152938843, "logits/rejected": -1.1151111125946045, "logps/chosen": -482.53857421875, "logps/rejected": -528.3353271484375, "loss": 0.4897, "rewards/accuracies": 0.75, "rewards/chosen": -2.206721544265747, "rewards/margins": 0.8017439842224121, "rewards/rejected": -3.008465528488159, "step": 630 }, { "epoch": 1.0, "step": 637, "total_flos": 0.0, "train_loss": 0.5397420258873861, "train_runtime": 14909.9591, "train_samples_per_second": 4.1, "train_steps_per_second": 0.043 } ], "logging_steps": 10, "max_steps": 637, "num_input_tokens_seen": 0, "num_train_epochs": 1, "save_steps": 100, "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 }