{ "best_metric": null, "best_model_checkpoint": null, "epoch": 0.9988571428571429, "eval_steps": 50, "global_step": 437, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.022857142857142857, "grad_norm": 5.484237277918217, "learning_rate": 1.1363636363636363e-07, "logits/chosen": -2.6541342735290527, "logits/rejected": -2.6346869468688965, "logps/chosen": -287.93011474609375, "logps/rejected": -266.5004577636719, "loss": 0.6931, "rewards/accuracies": 0.40625, "rewards/chosen": -0.00015118246665224433, "rewards/margins": -3.311875479994342e-05, "rewards/rejected": -0.00011806372640421614, "step": 10 }, { "epoch": 0.045714285714285714, "grad_norm": 5.141117524550972, "learning_rate": 2.2727272727272726e-07, "logits/chosen": -2.6553006172180176, "logits/rejected": -2.623598575592041, "logps/chosen": -271.3798828125, "logps/rejected": -260.2137145996094, "loss": 0.6925, "rewards/accuracies": 0.59375, "rewards/chosen": 0.003302179742604494, "rewards/margins": 0.0022685881704092026, "rewards/rejected": 0.0010335915721952915, "step": 20 }, { "epoch": 0.06857142857142857, "grad_norm": 4.867644605403788, "learning_rate": 3.4090909090909085e-07, "logits/chosen": -2.6496236324310303, "logits/rejected": -2.558840751647949, "logps/chosen": -271.86370849609375, "logps/rejected": -254.6623077392578, "loss": 0.6896, "rewards/accuracies": 0.71875, "rewards/chosen": 0.020251339301466942, "rewards/margins": 0.011756613850593567, "rewards/rejected": 0.008494723588228226, "step": 30 }, { "epoch": 0.09142857142857143, "grad_norm": 4.600987966838872, "learning_rate": 4.545454545454545e-07, "logits/chosen": -2.6127796173095703, "logits/rejected": -2.5982162952423096, "logps/chosen": -261.87078857421875, "logps/rejected": -255.64682006835938, "loss": 0.6811, "rewards/accuracies": 0.6937500238418579, "rewards/chosen": 0.039037298411130905, "rewards/margins": 0.04095984250307083, "rewards/rejected": -0.0019225452560931444, "step": 40 }, { "epoch": 0.11428571428571428, "grad_norm": 6.01325732118956, "learning_rate": 4.997124959943201e-07, "logits/chosen": -2.59413743019104, "logits/rejected": -2.525812864303589, "logps/chosen": -288.6096496582031, "logps/rejected": -264.43359375, "loss": 0.6661, "rewards/accuracies": 0.762499988079071, "rewards/chosen": 0.008406409062445164, "rewards/margins": 0.11031482368707657, "rewards/rejected": -0.10190840810537338, "step": 50 }, { "epoch": 0.11428571428571428, "eval_logits/chosen": -2.619272232055664, "eval_logits/rejected": -2.573228359222412, "eval_logps/chosen": -263.4992370605469, "eval_logps/rejected": -264.2141418457031, "eval_loss": 0.6540683507919312, "eval_rewards/accuracies": 0.6551724076271057, "eval_rewards/chosen": -0.0353790745139122, "eval_rewards/margins": 0.12399081885814667, "eval_rewards/rejected": -0.15936988592147827, "eval_runtime": 93.0218, "eval_samples_per_second": 19.684, "eval_steps_per_second": 0.312, "step": 50 }, { "epoch": 0.13714285714285715, "grad_norm": 8.643694975506126, "learning_rate": 4.979579212164186e-07, "logits/chosen": -2.6212775707244873, "logits/rejected": -2.5676815509796143, "logps/chosen": -305.444091796875, "logps/rejected": -272.07867431640625, "loss": 0.6429, "rewards/accuracies": 0.71875, "rewards/chosen": -0.05958025902509689, "rewards/margins": 0.1966249793767929, "rewards/rejected": -0.2562052309513092, "step": 60 }, { "epoch": 0.16, "grad_norm": 12.272483900335105, "learning_rate": 4.946196886175515e-07, "logits/chosen": -2.59287166595459, "logits/rejected": -2.5041251182556152, "logps/chosen": -321.8868103027344, "logps/rejected": -300.4162292480469, "loss": 0.629, "rewards/accuracies": 0.706250011920929, "rewards/chosen": -0.18531829118728638, "rewards/margins": 0.3314962685108185, "rewards/rejected": -0.5168145895004272, "step": 70 }, { "epoch": 0.18285714285714286, "grad_norm": 13.506330348124033, "learning_rate": 4.897191188239667e-07, "logits/chosen": -2.069486141204834, "logits/rejected": -1.9483461380004883, "logps/chosen": -318.3644714355469, "logps/rejected": -329.19659423828125, "loss": 0.6107, "rewards/accuracies": 0.637499988079071, "rewards/chosen": -0.5116854906082153, "rewards/margins": 0.31293898820877075, "rewards/rejected": -0.8246244192123413, "step": 80 }, { "epoch": 0.2057142857142857, "grad_norm": 12.553950954592443, "learning_rate": 4.832875107981763e-07, "logits/chosen": -1.1866068840026855, "logits/rejected": -0.9298005104064941, "logps/chosen": -316.0557556152344, "logps/rejected": -380.7778625488281, "loss": 0.5915, "rewards/accuracies": 0.706250011920929, "rewards/chosen": -0.7446193695068359, "rewards/margins": 0.5451987981796265, "rewards/rejected": -1.2898180484771729, "step": 90 }, { "epoch": 0.22857142857142856, "grad_norm": 30.03309840012713, "learning_rate": 4.753659419387223e-07, "logits/chosen": -0.9629250764846802, "logits/rejected": -0.5570683479309082, "logps/chosen": -331.9052429199219, "logps/rejected": -379.1312561035156, "loss": 0.561, "rewards/accuracies": 0.71875, "rewards/chosen": -0.7065539360046387, "rewards/margins": 0.7784632444381714, "rewards/rejected": -1.4850170612335205, "step": 100 }, { "epoch": 0.22857142857142856, "eval_logits/chosen": -0.4111216068267822, "eval_logits/rejected": -0.007617468014359474, "eval_logps/chosen": -365.1130676269531, "eval_logps/rejected": -431.9446716308594, "eval_loss": 0.5662747621536255, "eval_rewards/accuracies": 0.7068965435028076, "eval_rewards/chosen": -1.0515177249908447, "eval_rewards/margins": 0.7851568460464478, "eval_rewards/rejected": -1.8366745710372925, "eval_runtime": 93.7201, "eval_samples_per_second": 19.537, "eval_steps_per_second": 0.309, "step": 100 }, { "epoch": 0.25142857142857145, "grad_norm": 27.312780453008052, "learning_rate": 4.660050057270191e-07, "logits/chosen": -0.45199713110923767, "logits/rejected": 0.33052048087120056, "logps/chosen": -373.90545654296875, "logps/rejected": -399.00909423828125, "loss": 0.5415, "rewards/accuracies": 0.793749988079071, "rewards/chosen": -0.6764419674873352, "rewards/margins": 0.9693374633789062, "rewards/rejected": -1.6457793712615967, "step": 110 }, { "epoch": 0.2742857142857143, "grad_norm": 29.282942520454633, "learning_rate": 4.5526448859687144e-07, "logits/chosen": 0.3397678732872009, "logits/rejected": 0.8190152049064636, "logps/chosen": -406.60565185546875, "logps/rejected": -463.072021484375, "loss": 0.5753, "rewards/accuracies": 0.762499988079071, "rewards/chosen": -1.3377411365509033, "rewards/margins": 0.7959097623825073, "rewards/rejected": -2.133650779724121, "step": 120 }, { "epoch": 0.29714285714285715, "grad_norm": 24.36403749878471, "learning_rate": 4.432129880904388e-07, "logits/chosen": 0.4417182505130768, "logits/rejected": 1.1421695947647095, "logps/chosen": -371.94744873046875, "logps/rejected": -401.11962890625, "loss": 0.5483, "rewards/accuracies": 0.6187499761581421, "rewards/chosen": -0.9846571683883667, "rewards/margins": 0.6376141309738159, "rewards/rejected": -1.6222712993621826, "step": 130 }, { "epoch": 0.32, "grad_norm": 22.333886590434165, "learning_rate": 4.299274747394055e-07, "logits/chosen": 0.7880607843399048, "logits/rejected": 1.3198670148849487, "logps/chosen": -435.3951110839844, "logps/rejected": -469.02911376953125, "loss": 0.549, "rewards/accuracies": 0.731249988079071, "rewards/chosen": -1.175621747970581, "rewards/margins": 0.9118834733963013, "rewards/rejected": -2.0875051021575928, "step": 140 }, { "epoch": 0.34285714285714286, "grad_norm": 21.900935378631306, "learning_rate": 4.1549280046953653e-07, "logits/chosen": 1.040090560913086, "logits/rejected": 1.5702952146530151, "logps/chosen": -405.8429870605469, "logps/rejected": -457.5497131347656, "loss": 0.5324, "rewards/accuracies": 0.7437499761581421, "rewards/chosen": -1.0978893041610718, "rewards/margins": 0.8218156099319458, "rewards/rejected": -1.9197051525115967, "step": 150 }, { "epoch": 0.34285714285714286, "eval_logits/chosen": 1.325810432434082, "eval_logits/rejected": 2.0029168128967285, "eval_logps/chosen": -405.1368103027344, "eval_logps/rejected": -492.1299743652344, "eval_loss": 0.5436938405036926, "eval_rewards/accuracies": 0.6853448152542114, "eval_rewards/chosen": -1.451755166053772, "eval_rewards/margins": 0.986772894859314, "eval_rewards/rejected": -2.438527822494507, "eval_runtime": 94.6665, "eval_samples_per_second": 19.342, "eval_steps_per_second": 0.306, "step": 150 }, { "epoch": 0.3657142857142857, "grad_norm": 20.99267450356195, "learning_rate": 4.000011566683401e-07, "logits/chosen": 0.8327977061271667, "logits/rejected": 1.764451265335083, "logps/chosen": -447.289794921875, "logps/rejected": -494.8617248535156, "loss": 0.5254, "rewards/accuracies": 0.699999988079071, "rewards/chosen": -1.381363034248352, "rewards/margins": 0.8699455261230469, "rewards/rejected": -2.2513084411621094, "step": 160 }, { "epoch": 0.38857142857142857, "grad_norm": 19.625600234650467, "learning_rate": 3.8355148537705047e-07, "logits/chosen": 0.17654064297676086, "logits/rejected": 1.0507433414459229, "logps/chosen": -395.59124755859375, "logps/rejected": -480.58038330078125, "loss": 0.5315, "rewards/accuracies": 0.75, "rewards/chosen": -1.0189447402954102, "rewards/margins": 1.0480358600616455, "rewards/rejected": -2.0669806003570557, "step": 170 }, { "epoch": 0.4114285714285714, "grad_norm": 22.131134389700858, "learning_rate": 3.662488473675315e-07, "logits/chosen": 1.4423513412475586, "logits/rejected": 2.64449143409729, "logps/chosen": -448.270263671875, "logps/rejected": -503.065185546875, "loss": 0.5355, "rewards/accuracies": 0.75, "rewards/chosen": -1.4582993984222412, "rewards/margins": 1.1077758073806763, "rewards/rejected": -2.566075563430786, "step": 180 }, { "epoch": 0.4342857142857143, "grad_norm": 22.169020143018557, "learning_rate": 3.48203751140067e-07, "logits/chosen": 0.9658209085464478, "logits/rejected": 2.3707427978515625, "logps/chosen": -405.305419921875, "logps/rejected": -484.43499755859375, "loss": 0.538, "rewards/accuracies": 0.7437499761581421, "rewards/chosen": -1.4446977376937866, "rewards/margins": 1.050331711769104, "rewards/rejected": -2.4950294494628906, "step": 190 }, { "epoch": 0.45714285714285713, "grad_norm": 18.01149891090628, "learning_rate": 3.2953144712759537e-07, "logits/chosen": 1.7644532918930054, "logits/rejected": 2.6767263412475586, "logps/chosen": -395.5130310058594, "logps/rejected": -487.91192626953125, "loss": 0.5261, "rewards/accuracies": 0.737500011920929, "rewards/chosen": -1.451765537261963, "rewards/margins": 1.0395876169204712, "rewards/rejected": -2.4913532733917236, "step": 200 }, { "epoch": 0.45714285714285713, "eval_logits/chosen": 1.7313151359558105, "eval_logits/rejected": 2.7388670444488525, "eval_logps/chosen": -416.20770263671875, "eval_logps/rejected": -507.4054870605469, "eval_loss": 0.5247488021850586, "eval_rewards/accuracies": 0.6853448152542114, "eval_rewards/chosen": -1.5624639987945557, "eval_rewards/margins": 1.0288186073303223, "eval_rewards/rejected": -2.591282606124878, "eval_runtime": 94.5093, "eval_samples_per_second": 19.374, "eval_steps_per_second": 0.307, "step": 200 }, { "epoch": 0.48, "grad_norm": 19.57404608223085, "learning_rate": 3.103511916141658e-07, "logits/chosen": 1.504326581954956, "logits/rejected": 2.792241334915161, "logps/chosen": -433.87969970703125, "logps/rejected": -526.0897216796875, "loss": 0.517, "rewards/accuracies": 0.737500011920929, "rewards/chosen": -1.4803601503372192, "rewards/margins": 1.1599363088607788, "rewards/rejected": -2.640296459197998, "step": 210 }, { "epoch": 0.5028571428571429, "grad_norm": 21.692320362176854, "learning_rate": 2.9078548506882117e-07, "logits/chosen": 1.4797990322113037, "logits/rejected": 2.450709581375122, "logps/chosen": -449.7752990722656, "logps/rejected": -549.132080078125, "loss": 0.5296, "rewards/accuracies": 0.731249988079071, "rewards/chosen": -1.4539282321929932, "rewards/margins": 1.1360087394714355, "rewards/rejected": -2.5899369716644287, "step": 220 }, { "epoch": 0.5257142857142857, "grad_norm": 19.33385563287397, "learning_rate": 2.709592897595191e-07, "logits/chosen": 1.3153431415557861, "logits/rejected": 2.2327983379364014, "logps/chosen": -424.7271423339844, "logps/rejected": -506.617919921875, "loss": 0.5439, "rewards/accuracies": 0.706250011920929, "rewards/chosen": -1.3993257284164429, "rewards/margins": 0.8849741816520691, "rewards/rejected": -2.2843000888824463, "step": 230 }, { "epoch": 0.5485714285714286, "grad_norm": 19.794275801154928, "learning_rate": 2.509992316440332e-07, "logits/chosen": 2.2576842308044434, "logits/rejected": 3.4659297466278076, "logps/chosen": -488.4747619628906, "logps/rejected": -558.2715454101562, "loss": 0.5332, "rewards/accuracies": 0.793749988079071, "rewards/chosen": -1.8078649044036865, "rewards/margins": 1.1585441827774048, "rewards/rejected": -2.966409206390381, "step": 240 }, { "epoch": 0.5714285714285714, "grad_norm": 19.209773850229293, "learning_rate": 2.3103279163519918e-07, "logits/chosen": 1.9105958938598633, "logits/rejected": 2.3923182487487793, "logps/chosen": -476.888671875, "logps/rejected": -540.67724609375, "loss": 0.5274, "rewards/accuracies": 0.6312500238418579, "rewards/chosen": -1.9962562322616577, "rewards/margins": 0.6005629301071167, "rewards/rejected": -2.5968191623687744, "step": 250 }, { "epoch": 0.5714285714285714, "eval_logits/chosen": 1.0144059658050537, "eval_logits/rejected": 2.126599073410034, "eval_logps/chosen": -428.11065673828125, "eval_logps/rejected": -528.8191528320312, "eval_loss": 0.514776349067688, "eval_rewards/accuracies": 0.7155172228813171, "eval_rewards/chosen": -1.6814932823181152, "eval_rewards/margins": 1.123926043510437, "eval_rewards/rejected": -2.8054192066192627, "eval_runtime": 95.0873, "eval_samples_per_second": 19.256, "eval_steps_per_second": 0.305, "step": 250 }, { "epoch": 0.5942857142857143, "grad_norm": 21.990080244291757, "learning_rate": 2.1118749140573358e-07, "logits/chosen": 1.3521721363067627, "logits/rejected": 2.6136393547058105, "logps/chosen": -439.0127868652344, "logps/rejected": -476.6556701660156, "loss": 0.5238, "rewards/accuracies": 0.7124999761581421, "rewards/chosen": -1.6318817138671875, "rewards/margins": 0.9099860191345215, "rewards/rejected": -2.54186749458313, "step": 260 }, { "epoch": 0.6171428571428571, "grad_norm": 21.20271695558916, "learning_rate": 1.9159007893272703e-07, "logits/chosen": 1.4235649108886719, "logits/rejected": 2.510680913925171, "logps/chosen": -468.30975341796875, "logps/rejected": -496.22406005859375, "loss": 0.5192, "rewards/accuracies": 0.65625, "rewards/chosen": -1.8018286228179932, "rewards/margins": 0.6999757885932922, "rewards/rejected": -2.5018038749694824, "step": 270 }, { "epoch": 0.64, "grad_norm": 18.377021107490634, "learning_rate": 1.7236571898357766e-07, "logits/chosen": 1.396390676498413, "logits/rejected": 2.4561963081359863, "logps/chosen": -388.86273193359375, "logps/rejected": -491.08612060546875, "loss": 0.5273, "rewards/accuracies": 0.7124999761581421, "rewards/chosen": -1.3913507461547852, "rewards/margins": 1.1841250658035278, "rewards/rejected": -2.5754759311676025, "step": 280 }, { "epoch": 0.6628571428571428, "grad_norm": 23.869092599907706, "learning_rate": 1.5363719371356882e-07, "logits/chosen": 1.4888983964920044, "logits/rejected": 2.697493076324463, "logps/chosen": -438.34613037109375, "logps/rejected": -505.64398193359375, "loss": 0.5086, "rewards/accuracies": 0.737500011920929, "rewards/chosen": -1.5241912603378296, "rewards/margins": 1.000713586807251, "rewards/rejected": -2.524904727935791, "step": 290 }, { "epoch": 0.6857142857142857, "grad_norm": 25.027316619015263, "learning_rate": 1.3552411848071565e-07, "logits/chosen": 1.6033051013946533, "logits/rejected": 3.075709581375122, "logps/chosen": -483.4491271972656, "logps/rejected": -563.5081787109375, "loss": 0.5, "rewards/accuracies": 0.737500011920929, "rewards/chosen": -1.700990080833435, "rewards/margins": 1.2899569272994995, "rewards/rejected": -2.9909472465515137, "step": 300 }, { "epoch": 0.6857142857142857, "eval_logits/chosen": 1.5160114765167236, "eval_logits/rejected": 2.702754259109497, "eval_logps/chosen": -428.75518798828125, "eval_logps/rejected": -535.8169555664062, "eval_loss": 0.5077716112136841, "eval_rewards/accuracies": 0.7198275923728943, "eval_rewards/chosen": -1.6879388093948364, "eval_rewards/margins": 1.187458872795105, "eval_rewards/rejected": -2.8753974437713623, "eval_runtime": 93.8182, "eval_samples_per_second": 19.516, "eval_steps_per_second": 0.309, "step": 300 }, { "epoch": 0.7085714285714285, "grad_norm": 20.06935454117094, "learning_rate": 1.1814217788631473e-07, "logits/chosen": 1.097944974899292, "logits/rejected": 2.714242458343506, "logps/chosen": -464.17645263671875, "logps/rejected": -550.5721435546875, "loss": 0.512, "rewards/accuracies": 0.762499988079071, "rewards/chosen": -1.509181261062622, "rewards/margins": 1.3204662799835205, "rewards/rejected": -2.8296473026275635, "step": 310 }, { "epoch": 0.7314285714285714, "grad_norm": 20.37966135449331, "learning_rate": 1.0160238692045331e-07, "logits/chosen": 1.1582145690917969, "logits/rejected": 2.5812854766845703, "logps/chosen": -450.8458557128906, "logps/rejected": -533.5846557617188, "loss": 0.5203, "rewards/accuracies": 0.7749999761581421, "rewards/chosen": -1.5372800827026367, "rewards/margins": 1.0487934350967407, "rewards/rejected": -2.586073637008667, "step": 320 }, { "epoch": 0.7542857142857143, "grad_norm": 20.9552841041291, "learning_rate": 8.601038193139438e-08, "logits/chosen": 1.9143211841583252, "logits/rejected": 2.735372304916382, "logps/chosen": -400.63372802734375, "logps/rejected": -489.5486755371094, "loss": 0.52, "rewards/accuracies": 0.6937500238418579, "rewards/chosen": -1.5494529008865356, "rewards/margins": 0.9419981241226196, "rewards/rejected": -2.491450786590576, "step": 330 }, { "epoch": 0.7771428571428571, "grad_norm": 20.770737913391446, "learning_rate": 7.146574594727572e-08, "logits/chosen": 1.9478594064712524, "logits/rejected": 3.201336622238159, "logps/chosen": -425.11126708984375, "logps/rejected": -508.321533203125, "loss": 0.5043, "rewards/accuracies": 0.762499988079071, "rewards/chosen": -1.7910737991333008, "rewards/margins": 1.1562799215316772, "rewards/rejected": -2.947354316711426, "step": 340 }, { "epoch": 0.8, "grad_norm": 23.43196809715842, "learning_rate": 5.8061372659157306e-08, "logits/chosen": 2.0667195320129395, "logits/rejected": 3.518019914627075, "logps/chosen": -500.10174560546875, "logps/rejected": -575.167236328125, "loss": 0.4879, "rewards/accuracies": 0.737500011920929, "rewards/chosen": -1.8992862701416016, "rewards/margins": 1.168575406074524, "rewards/rejected": -3.067861557006836, "step": 350 }, { "epoch": 0.8, "eval_logits/chosen": 2.206516742706299, "eval_logits/rejected": 3.2477219104766846, "eval_logps/chosen": -448.67852783203125, "eval_logps/rejected": -555.7252197265625, "eval_loss": 0.5050091743469238, "eval_rewards/accuracies": 0.7198275923728943, "eval_rewards/chosen": -1.8871723413467407, "eval_rewards/margins": 1.187308430671692, "eval_rewards/rejected": -3.0744810104370117, "eval_runtime": 93.1425, "eval_samples_per_second": 19.658, "eval_steps_per_second": 0.311, "step": 350 }, { "epoch": 0.8228571428571428, "grad_norm": 21.572101434555105, "learning_rate": 4.5882873127531614e-08, "logits/chosen": 2.518174648284912, "logits/rejected": 3.5241615772247314, "logps/chosen": -472.5306701660156, "logps/rejected": -603.8782958984375, "loss": 0.5056, "rewards/accuracies": 0.8125, "rewards/chosen": -1.8001983165740967, "rewards/margins": 1.5649583339691162, "rewards/rejected": -3.365156650543213, "step": 360 }, { "epoch": 0.8457142857142858, "grad_norm": 24.32119336663692, "learning_rate": 3.500802900154412e-08, "logits/chosen": 2.2133302688598633, "logits/rejected": 2.970715284347534, "logps/chosen": -420.18707275390625, "logps/rejected": -547.6619873046875, "loss": 0.5034, "rewards/accuracies": 0.762499988079071, "rewards/chosen": -1.6418088674545288, "rewards/margins": 1.213688611984253, "rewards/rejected": -2.855497360229492, "step": 370 }, { "epoch": 0.8685714285714285, "grad_norm": 25.458784716444757, "learning_rate": 2.550629574310309e-08, "logits/chosen": 2.7819252014160156, "logits/rejected": 3.5777878761291504, "logps/chosen": -435.19268798828125, "logps/rejected": -533.6688842773438, "loss": 0.4987, "rewards/accuracies": 0.6812499761581421, "rewards/chosen": -1.7933107614517212, "rewards/margins": 0.9658328890800476, "rewards/rejected": -2.759143352508545, "step": 380 }, { "epoch": 0.8914285714285715, "grad_norm": 20.416342139746472, "learning_rate": 1.7438359028687983e-08, "logits/chosen": 2.2803592681884766, "logits/rejected": 3.287682056427002, "logps/chosen": -436.3902282714844, "logps/rejected": -546.454833984375, "loss": 0.5096, "rewards/accuracies": 0.6937500238418579, "rewards/chosen": -1.6981942653656006, "rewards/margins": 1.2526156902313232, "rewards/rejected": -2.950809955596924, "step": 390 }, { "epoch": 0.9142857142857143, "grad_norm": 23.31604206083898, "learning_rate": 1.0855747162029361e-08, "logits/chosen": 1.8559446334838867, "logits/rejected": 2.6898691654205322, "logps/chosen": -443.5389709472656, "logps/rejected": -552.5650634765625, "loss": 0.5082, "rewards/accuracies": 0.75, "rewards/chosen": -1.6940586566925049, "rewards/margins": 1.158569574356079, "rewards/rejected": -2.852627992630005, "step": 400 }, { "epoch": 0.9142857142857143, "eval_logits/chosen": 1.9658149480819702, "eval_logits/rejected": 3.059574842453003, "eval_logps/chosen": -435.48773193359375, "eval_logps/rejected": -548.0841064453125, "eval_loss": 0.5012447834014893, "eval_rewards/accuracies": 0.7198275923728943, "eval_rewards/chosen": -1.7552645206451416, "eval_rewards/margins": 1.2428052425384521, "eval_rewards/rejected": -2.9980695247650146, "eval_runtime": 94.0643, "eval_samples_per_second": 19.465, "eval_steps_per_second": 0.308, "step": 400 }, { "epoch": 0.9371428571428572, "grad_norm": 19.227803698254093, "learning_rate": 5.8005019731033615e-09, "logits/chosen": 2.1140646934509277, "logits/rejected": 3.0444817543029785, "logps/chosen": -424.0716247558594, "logps/rejected": -529.0479736328125, "loss": 0.4899, "rewards/accuracies": 0.768750011920929, "rewards/chosen": -1.6376755237579346, "rewards/margins": 1.076155662536621, "rewards/rejected": -2.7138311862945557, "step": 410 }, { "epoch": 0.96, "grad_norm": 19.178542640219938, "learning_rate": 2.3049103053431886e-09, "logits/chosen": 1.7973829507827759, "logits/rejected": 2.994089365005493, "logps/chosen": -480.82342529296875, "logps/rejected": -606.6632690429688, "loss": 0.5082, "rewards/accuracies": 0.737500011920929, "rewards/chosen": -1.7015644311904907, "rewards/margins": 1.422523856163025, "rewards/rejected": -3.1240882873535156, "step": 420 }, { "epoch": 0.9828571428571429, "grad_norm": 23.036308488754496, "learning_rate": 3.9129780600541397e-10, "logits/chosen": 2.1120381355285645, "logits/rejected": 3.065523862838745, "logps/chosen": -451.97515869140625, "logps/rejected": -554.6312866210938, "loss": 0.4855, "rewards/accuracies": 0.78125, "rewards/chosen": -1.6638119220733643, "rewards/margins": 1.303351640701294, "rewards/rejected": -2.967163324356079, "step": 430 }, { "epoch": 0.9988571428571429, "step": 437, "total_flos": 0.0, "train_loss": 0.5499054316252265, "train_runtime": 11738.6737, "train_samples_per_second": 4.77, "train_steps_per_second": 0.037 } ], "logging_steps": 10, "max_steps": 437, "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": 8, "trial_name": null, "trial_params": null }