{ "best_metric": null, "best_model_checkpoint": null, "epoch": 0.9989071038251366, "eval_steps": 400, "global_step": 457, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.01092896174863388, "grad_norm": 334.04762282844746, "learning_rate": 1.0869565217391303e-07, "logits/chosen": -1.0183193683624268, "logits/rejected": -1.0128684043884277, "logps/chosen": -0.2805940508842468, "logps/rejected": -0.28598836064338684, "loss": 3.2497, "rewards/accuracies": 0.53125, "rewards/chosen": -2.805940628051758, "rewards/margins": 0.05394287034869194, "rewards/rejected": -2.8598835468292236, "step": 5 }, { "epoch": 0.02185792349726776, "grad_norm": 180.71056791479407, "learning_rate": 2.1739130434782607e-07, "logits/chosen": -1.0511164665222168, "logits/rejected": -1.0003410577774048, "logps/chosen": -0.2573581039905548, "logps/rejected": -0.27061566710472107, "loss": 3.257, "rewards/accuracies": 0.53125, "rewards/chosen": -2.5735812187194824, "rewards/margins": 0.13257600367069244, "rewards/rejected": -2.7061567306518555, "step": 10 }, { "epoch": 0.03278688524590164, "grad_norm": 223.3194766304879, "learning_rate": 3.260869565217391e-07, "logits/chosen": -0.9949806928634644, "logits/rejected": -0.9507344961166382, "logps/chosen": -0.26757049560546875, "logps/rejected": -0.2733513414859772, "loss": 3.2029, "rewards/accuracies": 0.518750011920929, "rewards/chosen": -2.6757049560546875, "rewards/margins": 0.05780839920043945, "rewards/rejected": -2.733513355255127, "step": 15 }, { "epoch": 0.04371584699453552, "grad_norm": 208.56574432084847, "learning_rate": 4.3478260869565214e-07, "logits/chosen": -0.9485995173454285, "logits/rejected": -0.8978453874588013, "logps/chosen": -0.2727873921394348, "logps/rejected": -0.28461021184921265, "loss": 3.2536, "rewards/accuracies": 0.550000011920929, "rewards/chosen": -2.7278740406036377, "rewards/margins": 0.11822839826345444, "rewards/rejected": -2.846102237701416, "step": 20 }, { "epoch": 0.0546448087431694, "grad_norm": 273.9157814428763, "learning_rate": 5.434782608695652e-07, "logits/chosen": -0.9404302835464478, "logits/rejected": -0.8673273324966431, "logps/chosen": -0.2763015627861023, "logps/rejected": -0.29439201951026917, "loss": 3.1849, "rewards/accuracies": 0.581250011920929, "rewards/chosen": -2.7630152702331543, "rewards/margins": 0.18090477585792542, "rewards/rejected": -2.943920135498047, "step": 25 }, { "epoch": 0.06557377049180328, "grad_norm": 252.26801160371912, "learning_rate": 6.521739130434782e-07, "logits/chosen": -1.0567680597305298, "logits/rejected": -0.9916139841079712, "logps/chosen": -0.26832228899002075, "logps/rejected": -0.2841094732284546, "loss": 3.2649, "rewards/accuracies": 0.5062500238418579, "rewards/chosen": -2.683222770690918, "rewards/margins": 0.15787188708782196, "rewards/rejected": -2.841094493865967, "step": 30 }, { "epoch": 0.07650273224043716, "grad_norm": 120.68425546239968, "learning_rate": 7.608695652173913e-07, "logits/chosen": -1.0046221017837524, "logits/rejected": -0.9383124113082886, "logps/chosen": -0.2608323395252228, "logps/rejected": -0.27886277437210083, "loss": 3.1693, "rewards/accuracies": 0.48750001192092896, "rewards/chosen": -2.608323335647583, "rewards/margins": 0.1803044080734253, "rewards/rejected": -2.788627862930298, "step": 35 }, { "epoch": 0.08743169398907104, "grad_norm": 146.20382704567854, "learning_rate": 8.695652173913043e-07, "logits/chosen": -0.9549115896224976, "logits/rejected": -0.8933135271072388, "logps/chosen": -0.28391939401626587, "logps/rejected": -0.29877811670303345, "loss": 3.2899, "rewards/accuracies": 0.518750011920929, "rewards/chosen": -2.83919358253479, "rewards/margins": 0.14858722686767578, "rewards/rejected": -2.987781047821045, "step": 40 }, { "epoch": 0.09836065573770492, "grad_norm": 93.83055902381597, "learning_rate": 9.782608695652173e-07, "logits/chosen": -1.0223612785339355, "logits/rejected": -0.9385920763015747, "logps/chosen": -0.2871010899543762, "logps/rejected": -0.3065423369407654, "loss": 3.0292, "rewards/accuracies": 0.5062500238418579, "rewards/chosen": -2.8710105419158936, "rewards/margins": 0.19441291689872742, "rewards/rejected": -3.0654234886169434, "step": 45 }, { "epoch": 0.1092896174863388, "grad_norm": 281.2686273771578, "learning_rate": 9.997663088532014e-07, "logits/chosen": -0.9519433975219727, "logits/rejected": -0.8702558279037476, "logps/chosen": -0.28290504217147827, "logps/rejected": -0.29115915298461914, "loss": 3.1626, "rewards/accuracies": 0.48750001192092896, "rewards/chosen": -2.829050302505493, "rewards/margins": 0.08254138380289078, "rewards/rejected": -2.9115920066833496, "step": 50 }, { "epoch": 0.12021857923497267, "grad_norm": 121.70775424511473, "learning_rate": 9.98817312944725e-07, "logits/chosen": -0.9872657060623169, "logits/rejected": -0.8688336610794067, "logps/chosen": -0.2818766236305237, "logps/rejected": -0.3157486021518707, "loss": 3.0197, "rewards/accuracies": 0.543749988079071, "rewards/chosen": -2.8187661170959473, "rewards/margins": 0.3387202322483063, "rewards/rejected": -3.1574864387512207, "step": 55 }, { "epoch": 0.13114754098360656, "grad_norm": 86.612813297471, "learning_rate": 9.971397915250336e-07, "logits/chosen": -1.0298466682434082, "logits/rejected": -0.9812299609184265, "logps/chosen": -0.2719958424568176, "logps/rejected": -0.31191515922546387, "loss": 2.8695, "rewards/accuracies": 0.59375, "rewards/chosen": -2.719958782196045, "rewards/margins": 0.399192750453949, "rewards/rejected": -3.1191515922546387, "step": 60 }, { "epoch": 0.14207650273224043, "grad_norm": 161.66395420923772, "learning_rate": 9.94736194623663e-07, "logits/chosen": -0.9994015693664551, "logits/rejected": -0.9284494519233704, "logps/chosen": -0.31242281198501587, "logps/rejected": -0.34273454546928406, "loss": 3.0819, "rewards/accuracies": 0.53125, "rewards/chosen": -3.124228000640869, "rewards/margins": 0.30311769247055054, "rewards/rejected": -3.4273457527160645, "step": 65 }, { "epoch": 0.15300546448087432, "grad_norm": 335.25942540938917, "learning_rate": 9.916100327075037e-07, "logits/chosen": -0.9326862096786499, "logits/rejected": -0.9134162664413452, "logps/chosen": -0.3001170754432678, "logps/rejected": -0.3256034851074219, "loss": 2.8894, "rewards/accuracies": 0.581250011920929, "rewards/chosen": -3.0011706352233887, "rewards/margins": 0.2548643946647644, "rewards/rejected": -3.2560348510742188, "step": 70 }, { "epoch": 0.16393442622950818, "grad_norm": 79.67001422366646, "learning_rate": 9.877658715537428e-07, "logits/chosen": -0.9303417205810547, "logits/rejected": -0.9135540127754211, "logps/chosen": -0.32335883378982544, "logps/rejected": -0.35447338223457336, "loss": 2.9985, "rewards/accuracies": 0.625, "rewards/chosen": -3.233588695526123, "rewards/margins": 0.3111455738544464, "rewards/rejected": -3.544733762741089, "step": 75 }, { "epoch": 0.17486338797814208, "grad_norm": 83.46864509697639, "learning_rate": 9.832093255815216e-07, "logits/chosen": -0.942563533782959, "logits/rejected": -0.8771953582763672, "logps/chosen": -0.31826624274253845, "logps/rejected": -0.3393171727657318, "loss": 2.9663, "rewards/accuracies": 0.606249988079071, "rewards/chosen": -3.1826624870300293, "rewards/margins": 0.21050938963890076, "rewards/rejected": -3.393172025680542, "step": 80 }, { "epoch": 0.18579234972677597, "grad_norm": 119.60039352898374, "learning_rate": 9.779470496520441e-07, "logits/chosen": -0.9433110952377319, "logits/rejected": -0.8911749124526978, "logps/chosen": -0.31765812635421753, "logps/rejected": -0.3729430139064789, "loss": 2.8342, "rewards/accuracies": 0.625, "rewards/chosen": -3.176581621170044, "rewards/margins": 0.5528485178947449, "rewards/rejected": -3.7294304370880127, "step": 85 }, { "epoch": 0.19672131147540983, "grad_norm": 96.2051979280833, "learning_rate": 9.719867293491144e-07, "logits/chosen": -1.0151774883270264, "logits/rejected": -0.9306642413139343, "logps/chosen": -0.3450705409049988, "logps/rejected": -0.38217830657958984, "loss": 2.9117, "rewards/accuracies": 0.5562499761581421, "rewards/chosen": -3.4507057666778564, "rewards/margins": 0.37107738852500916, "rewards/rejected": -3.8217825889587402, "step": 90 }, { "epoch": 0.20765027322404372, "grad_norm": 90.0031374445934, "learning_rate": 9.653370697542987e-07, "logits/chosen": -0.9307557344436646, "logits/rejected": -0.9305530786514282, "logps/chosen": -0.34383976459503174, "logps/rejected": -0.3695612847805023, "loss": 2.7261, "rewards/accuracies": 0.581250011920929, "rewards/chosen": -3.4383978843688965, "rewards/margins": 0.2572151720523834, "rewards/rejected": -3.695613145828247, "step": 95 }, { "epoch": 0.2185792349726776, "grad_norm": 106.07682375515387, "learning_rate": 9.580077827331037e-07, "logits/chosen": -0.9448221921920776, "logits/rejected": -0.8973434567451477, "logps/chosen": -0.38050413131713867, "logps/rejected": -0.4358042776584625, "loss": 2.7135, "rewards/accuracies": 0.637499988079071, "rewards/chosen": -3.805041551589966, "rewards/margins": 0.5530012249946594, "rewards/rejected": -4.3580427169799805, "step": 100 }, { "epoch": 0.22950819672131148, "grad_norm": 124.64829049828732, "learning_rate": 9.500095727507419e-07, "logits/chosen": -0.9876478910446167, "logits/rejected": -0.9548322558403015, "logps/chosen": -0.3621714115142822, "logps/rejected": -0.4053111672401428, "loss": 2.8156, "rewards/accuracies": 0.625, "rewards/chosen": -3.6217141151428223, "rewards/margins": 0.43139734864234924, "rewards/rejected": -4.0531110763549805, "step": 105 }, { "epoch": 0.24043715846994534, "grad_norm": 109.72879050820204, "learning_rate": 9.413541212382004e-07, "logits/chosen": -0.9878107905387878, "logits/rejected": -0.9694309234619141, "logps/chosen": -0.37187474966049194, "logps/rejected": -0.4513128697872162, "loss": 2.6674, "rewards/accuracies": 0.6499999761581421, "rewards/chosen": -3.718747615814209, "rewards/margins": 0.7943812608718872, "rewards/rejected": -4.513128280639648, "step": 110 }, { "epoch": 0.25136612021857924, "grad_norm": 89.69262563371302, "learning_rate": 9.320540695314438e-07, "logits/chosen": -0.9974454045295715, "logits/rejected": -0.9548517465591431, "logps/chosen": -0.3737660050392151, "logps/rejected": -0.47232741117477417, "loss": 2.5733, "rewards/accuracies": 0.6937500238418579, "rewards/chosen": -3.7376599311828613, "rewards/margins": 0.9856147766113281, "rewards/rejected": -4.7232747077941895, "step": 115 }, { "epoch": 0.26229508196721313, "grad_norm": 82.76010396371875, "learning_rate": 9.221230004086721e-07, "logits/chosen": -1.0242812633514404, "logits/rejected": -0.949226975440979, "logps/chosen": -0.3816962242126465, "logps/rejected": -0.4385160803794861, "loss": 2.6795, "rewards/accuracies": 0.6812499761581421, "rewards/chosen": -3.816962480545044, "rewards/margins": 0.5681988000869751, "rewards/rejected": -4.38516092300415, "step": 120 }, { "epoch": 0.273224043715847, "grad_norm": 186.60007457771528, "learning_rate": 9.11575418252596e-07, "logits/chosen": -0.9328588247299194, "logits/rejected": -0.8940401077270508, "logps/chosen": -0.395950585603714, "logps/rejected": -0.46849578619003296, "loss": 2.5289, "rewards/accuracies": 0.6875, "rewards/chosen": -3.9595062732696533, "rewards/margins": 0.7254521250724792, "rewards/rejected": -4.684957981109619, "step": 125 }, { "epoch": 0.28415300546448086, "grad_norm": 91.04741453080123, "learning_rate": 9.004267278667031e-07, "logits/chosen": -0.9401098489761353, "logits/rejected": -0.9308618307113647, "logps/chosen": -0.41634416580200195, "logps/rejected": -0.5375351905822754, "loss": 2.5277, "rewards/accuracies": 0.7124999761581421, "rewards/chosen": -4.1634416580200195, "rewards/margins": 1.2119100093841553, "rewards/rejected": -5.375351905822754, "step": 130 }, { "epoch": 0.29508196721311475, "grad_norm": 72.34007677315606, "learning_rate": 8.886932119764565e-07, "logits/chosen": -0.9937323331832886, "logits/rejected": -0.9037588834762573, "logps/chosen": -0.39853519201278687, "logps/rejected": -0.49872303009033203, "loss": 2.4319, "rewards/accuracies": 0.6625000238418579, "rewards/chosen": -3.9853522777557373, "rewards/margins": 1.0018781423568726, "rewards/rejected": -4.987229824066162, "step": 135 }, { "epoch": 0.30601092896174864, "grad_norm": 83.52757366186898, "learning_rate": 8.763920074482809e-07, "logits/chosen": -0.9973970651626587, "logits/rejected": -0.9379005432128906, "logps/chosen": -0.42847341299057007, "logps/rejected": -0.560250461101532, "loss": 2.2317, "rewards/accuracies": 0.706250011920929, "rewards/chosen": -4.28473424911499, "rewards/margins": 1.3177704811096191, "rewards/rejected": -5.602505207061768, "step": 140 }, { "epoch": 0.31693989071038253, "grad_norm": 106.73407366387536, "learning_rate": 8.635410802610723e-07, "logits/chosen": -0.9793792963027954, "logits/rejected": -0.9569779634475708, "logps/chosen": -0.4017811417579651, "logps/rejected": -0.4612775444984436, "loss": 2.3643, "rewards/accuracies": 0.7250000238418579, "rewards/chosen": -4.017810821533203, "rewards/margins": 0.5949639081954956, "rewards/rejected": -4.612774848937988, "step": 145 }, { "epoch": 0.32786885245901637, "grad_norm": 60.51485768923878, "learning_rate": 8.501591992667849e-07, "logits/chosen": -1.044575572013855, "logits/rejected": -1.0059934854507446, "logps/chosen": -0.4378751218318939, "logps/rejected": -0.5975905656814575, "loss": 2.2151, "rewards/accuracies": 0.737500011920929, "rewards/chosen": -4.378750801086426, "rewards/margins": 1.5971543788909912, "rewards/rejected": -5.975905418395996, "step": 150 }, { "epoch": 0.33879781420765026, "grad_norm": 68.73684458948155, "learning_rate": 8.362659087784152e-07, "logits/chosen": -0.9929186105728149, "logits/rejected": -0.9317314028739929, "logps/chosen": -0.4304039478302002, "logps/rejected": -0.5334851741790771, "loss": 2.3197, "rewards/accuracies": 0.675000011920929, "rewards/chosen": -4.304039478302002, "rewards/margins": 1.0308125019073486, "rewards/rejected": -5.3348517417907715, "step": 155 }, { "epoch": 0.34972677595628415, "grad_norm": 90.32236437312078, "learning_rate": 8.218815000254231e-07, "logits/chosen": -1.0294911861419678, "logits/rejected": -0.9657951593399048, "logps/chosen": -0.4845092296600342, "logps/rejected": -0.566204309463501, "loss": 2.3671, "rewards/accuracies": 0.6875, "rewards/chosen": -4.845091819763184, "rewards/margins": 0.81695157289505, "rewards/rejected": -5.662043571472168, "step": 160 }, { "epoch": 0.36065573770491804, "grad_norm": 83.55321323451464, "learning_rate": 8.07026981518276e-07, "logits/chosen": -1.0140776634216309, "logits/rejected": -0.9537431597709656, "logps/chosen": -0.46461719274520874, "logps/rejected": -0.5585328340530396, "loss": 2.3255, "rewards/accuracies": 0.737500011920929, "rewards/chosen": -4.646172046661377, "rewards/margins": 0.9391565322875977, "rewards/rejected": -5.585328578948975, "step": 165 }, { "epoch": 0.37158469945355194, "grad_norm": 91.94544798993923, "learning_rate": 7.917240483654e-07, "logits/chosen": -1.0053484439849854, "logits/rejected": -0.9363968968391418, "logps/chosen": -0.4600796699523926, "logps/rejected": -0.5375440716743469, "loss": 2.416, "rewards/accuracies": 0.6875, "rewards/chosen": -4.600796699523926, "rewards/margins": 0.7746434211730957, "rewards/rejected": -5.3754401206970215, "step": 170 }, { "epoch": 0.3825136612021858, "grad_norm": 91.07280013110446, "learning_rate": 7.759950505873521e-07, "logits/chosen": -1.0666639804840088, "logits/rejected": -1.029950737953186, "logps/chosen": -0.4774397313594818, "logps/rejected": -0.5452654957771301, "loss": 2.2653, "rewards/accuracies": 0.643750011920929, "rewards/chosen": -4.774397850036621, "rewards/margins": 0.6782568097114563, "rewards/rejected": -5.452654838562012, "step": 175 }, { "epoch": 0.39344262295081966, "grad_norm": 85.29762605205215, "learning_rate": 7.598629604744872e-07, "logits/chosen": -1.0595664978027344, "logits/rejected": -1.0487810373306274, "logps/chosen": -0.4570770263671875, "logps/rejected": -0.5979724526405334, "loss": 2.1524, "rewards/accuracies": 0.7875000238418579, "rewards/chosen": -4.570770263671875, "rewards/margins": 1.4089539051055908, "rewards/rejected": -5.979724407196045, "step": 180 }, { "epoch": 0.40437158469945356, "grad_norm": 67.2351189385839, "learning_rate": 7.433513390357989e-07, "logits/chosen": -1.0875113010406494, "logits/rejected": -1.0918848514556885, "logps/chosen": -0.4790703356266022, "logps/rejected": -0.6097405552864075, "loss": 2.164, "rewards/accuracies": 0.793749988079071, "rewards/chosen": -4.790703773498535, "rewards/margins": 1.3067026138305664, "rewards/rejected": -6.097405433654785, "step": 185 }, { "epoch": 0.41530054644808745, "grad_norm": 106.92479621624057, "learning_rate": 7.264843015879321e-07, "logits/chosen": -1.0997188091278076, "logits/rejected": -1.0514479875564575, "logps/chosen": -0.467475026845932, "logps/rejected": -0.6209039092063904, "loss": 2.2151, "rewards/accuracies": 0.7562500238418579, "rewards/chosen": -4.674751281738281, "rewards/margins": 1.5342884063720703, "rewards/rejected": -6.209038734436035, "step": 190 }, { "epoch": 0.4262295081967213, "grad_norm": 79.4429028708432, "learning_rate": 7.092864825346266e-07, "logits/chosen": -1.12069833278656, "logits/rejected": -1.0915439128875732, "logps/chosen": -0.5538396239280701, "logps/rejected": -0.7245546579360962, "loss": 2.2765, "rewards/accuracies": 0.737500011920929, "rewards/chosen": -5.538395881652832, "rewards/margins": 1.7071504592895508, "rewards/rejected": -7.245546817779541, "step": 195 }, { "epoch": 0.4371584699453552, "grad_norm": 79.46289075402596, "learning_rate": 6.917829993880302e-07, "logits/chosen": -1.1029081344604492, "logits/rejected": -1.0124149322509766, "logps/chosen": -0.5200821161270142, "logps/rejected": -0.6488970518112183, "loss": 2.1257, "rewards/accuracies": 0.762499988079071, "rewards/chosen": -5.200821399688721, "rewards/margins": 1.288149118423462, "rewards/rejected": -6.4889702796936035, "step": 200 }, { "epoch": 0.44808743169398907, "grad_norm": 67.1583967725435, "learning_rate": 6.739994160844309e-07, "logits/chosen": -1.0795233249664307, "logits/rejected": -1.0885263681411743, "logps/chosen": -0.48235517740249634, "logps/rejected": -0.6249179244041443, "loss": 2.0323, "rewards/accuracies": 0.7875000238418579, "rewards/chosen": -4.823552131652832, "rewards/margins": 1.4256280660629272, "rewards/rejected": -6.249179363250732, "step": 205 }, { "epoch": 0.45901639344262296, "grad_norm": 131.31523382011193, "learning_rate": 6.559617056479827e-07, "logits/chosen": -1.0998046398162842, "logits/rejected": -1.0920783281326294, "logps/chosen": -0.5168774724006653, "logps/rejected": -0.6893444657325745, "loss": 2.0968, "rewards/accuracies": 0.78125, "rewards/chosen": -5.168774127960205, "rewards/margins": 1.7246698141098022, "rewards/rejected": -6.893445014953613, "step": 210 }, { "epoch": 0.46994535519125685, "grad_norm": 92.8827942851212, "learning_rate": 6.376962122569567e-07, "logits/chosen": -1.1027876138687134, "logits/rejected": -1.0451091527938843, "logps/chosen": -0.5266443490982056, "logps/rejected": -0.6980344653129578, "loss": 2.2334, "rewards/accuracies": 0.7875000238418579, "rewards/chosen": -5.266442775726318, "rewards/margins": 1.7139012813568115, "rewards/rejected": -6.980344295501709, "step": 215 }, { "epoch": 0.4808743169398907, "grad_norm": 75.78200124794627, "learning_rate": 6.192296127679192e-07, "logits/chosen": -1.163720965385437, "logits/rejected": -1.1323108673095703, "logps/chosen": -0.5482050180435181, "logps/rejected": -0.6795030832290649, "loss": 2.0812, "rewards/accuracies": 0.7250000238418579, "rewards/chosen": -5.482050895690918, "rewards/margins": 1.3129799365997314, "rewards/rejected": -6.7950310707092285, "step": 220 }, { "epoch": 0.4918032786885246, "grad_norm": 68.50625090212893, "learning_rate": 6.005888777540319e-07, "logits/chosen": -1.1991808414459229, "logits/rejected": -1.1517974138259888, "logps/chosen": -0.5413341522216797, "logps/rejected": -0.6963613033294678, "loss": 2.1219, "rewards/accuracies": 0.7562500238418579, "rewards/chosen": -5.413341522216797, "rewards/margins": 1.5502712726593018, "rewards/rejected": -6.963613033294678, "step": 225 }, { "epoch": 0.5027322404371585, "grad_norm": 73.44567710043167, "learning_rate": 5.818012321143773e-07, "logits/chosen": -1.0955681800842285, "logits/rejected": -1.0907937288284302, "logps/chosen": -0.5347357988357544, "logps/rejected": -0.7202361226081848, "loss": 2.0511, "rewards/accuracies": 0.8062499761581421, "rewards/chosen": -5.347357749938965, "rewards/margins": 1.855002760887146, "rewards/rejected": -7.2023606300354, "step": 230 }, { "epoch": 0.5136612021857924, "grad_norm": 84.7486173764228, "learning_rate": 5.628941153118388e-07, "logits/chosen": -1.113163709640503, "logits/rejected": -1.0674149990081787, "logps/chosen": -0.534842312335968, "logps/rejected": -0.6825965046882629, "loss": 2.0329, "rewards/accuracies": 0.78125, "rewards/chosen": -5.348422527313232, "rewards/margins": 1.4775424003601074, "rewards/rejected": -6.825964450836182, "step": 235 }, { "epoch": 0.5245901639344263, "grad_norm": 70.6295555021999, "learning_rate": 5.438951412976098e-07, "logits/chosen": -1.1475393772125244, "logits/rejected": -1.1534240245819092, "logps/chosen": -0.5095857381820679, "logps/rejected": -0.6976131200790405, "loss": 1.8823, "rewards/accuracies": 0.7875000238418579, "rewards/chosen": -5.095856666564941, "rewards/margins": 1.8802751302719116, "rewards/rejected": -6.976132392883301, "step": 240 }, { "epoch": 0.5355191256830601, "grad_norm": 69.47924677139383, "learning_rate": 5.248320581808619e-07, "logits/chosen": -1.0735324621200562, "logits/rejected": -1.0195956230163574, "logps/chosen": -0.5255709290504456, "logps/rejected": -0.7123461961746216, "loss": 1.9839, "rewards/accuracies": 0.7437499761581421, "rewards/chosen": -5.255709648132324, "rewards/margins": 1.8677523136138916, "rewards/rejected": -7.123461723327637, "step": 245 }, { "epoch": 0.546448087431694, "grad_norm": 66.46597890028497, "learning_rate": 5.057327077024744e-07, "logits/chosen": -1.1226236820220947, "logits/rejected": -1.0827569961547852, "logps/chosen": -0.5162156224250793, "logps/rejected": -0.6480467319488525, "loss": 2.0699, "rewards/accuracies": 0.7250000238418579, "rewards/chosen": -5.162156105041504, "rewards/margins": 1.3183108568191528, "rewards/rejected": -6.480466365814209, "step": 250 }, { "epoch": 0.5573770491803278, "grad_norm": 76.19586118186004, "learning_rate": 4.866249845720132e-07, "logits/chosen": -1.107912540435791, "logits/rejected": -1.071297287940979, "logps/chosen": -0.5693866014480591, "logps/rejected": -0.735679566860199, "loss": 1.9599, "rewards/accuracies": 0.768750011920929, "rewards/chosen": -5.693865776062012, "rewards/margins": 1.6629297733306885, "rewards/rejected": -7.356795310974121, "step": 255 }, { "epoch": 0.5683060109289617, "grad_norm": 73.20621943329614, "learning_rate": 4.675367957273505e-07, "logits/chosen": -1.0760225057601929, "logits/rejected": -1.0634446144104004, "logps/chosen": -0.5194670557975769, "logps/rejected": -0.6913676261901855, "loss": 1.9788, "rewards/accuracies": 0.800000011920929, "rewards/chosen": -5.1946702003479, "rewards/margins": 1.7190055847167969, "rewards/rejected": -6.9136762619018555, "step": 260 }, { "epoch": 0.5792349726775956, "grad_norm": 111.34081159607314, "learning_rate": 4.4849601957642285e-07, "logits/chosen": -1.1172699928283691, "logits/rejected": -1.0767419338226318, "logps/chosen": -0.5310412645339966, "logps/rejected": -0.6969484686851501, "loss": 2.0084, "rewards/accuracies": 0.7562500238418579, "rewards/chosen": -5.310412406921387, "rewards/margins": 1.6590715646743774, "rewards/rejected": -6.969484806060791, "step": 265 }, { "epoch": 0.5901639344262295, "grad_norm": 79.77425345162416, "learning_rate": 4.295304652806592e-07, "logits/chosen": -1.1290314197540283, "logits/rejected": -1.098478078842163, "logps/chosen": -0.5287065505981445, "logps/rejected": -0.7144958972930908, "loss": 1.8771, "rewards/accuracies": 0.8062499761581421, "rewards/chosen": -5.2870659828186035, "rewards/margins": 1.8578933477401733, "rewards/rejected": -7.144959449768066, "step": 270 }, { "epoch": 0.6010928961748634, "grad_norm": 63.79079426666398, "learning_rate": 4.106678321395433e-07, "logits/chosen": -1.100162148475647, "logits/rejected": -1.0285935401916504, "logps/chosen": -0.5441292524337769, "logps/rejected": -0.6436029672622681, "loss": 1.9928, "rewards/accuracies": 0.65625, "rewards/chosen": -5.4412922859191895, "rewards/margins": 0.9947379231452942, "rewards/rejected": -6.43602991104126, "step": 275 }, { "epoch": 0.6120218579234973, "grad_norm": 84.61811881052614, "learning_rate": 3.9193566913562915e-07, "logits/chosen": -1.055173635482788, "logits/rejected": -1.0575199127197266, "logps/chosen": -0.5374099612236023, "logps/rejected": -0.7436326742172241, "loss": 1.9752, "rewards/accuracies": 0.737500011920929, "rewards/chosen": -5.374099254608154, "rewards/margins": 2.062227249145508, "rewards/rejected": -7.436326503753662, "step": 280 }, { "epoch": 0.6229508196721312, "grad_norm": 69.51995916576952, "learning_rate": 3.7336133469909623e-07, "logits/chosen": -1.1819168329238892, "logits/rejected": -1.1467748880386353, "logps/chosen": -0.5230523347854614, "logps/rejected": -0.7193282842636108, "loss": 1.923, "rewards/accuracies": 0.8062499761581421, "rewards/chosen": -5.230523109436035, "rewards/margins": 1.9627596139907837, "rewards/rejected": -7.1932830810546875, "step": 285 }, { "epoch": 0.6338797814207651, "grad_norm": 98.48874799948564, "learning_rate": 3.549719567506076e-07, "logits/chosen": -1.1354302167892456, "logits/rejected": -1.0900018215179443, "logps/chosen": -0.5486642718315125, "logps/rejected": -0.7162230610847473, "loss": 1.9529, "rewards/accuracies": 0.793749988079071, "rewards/chosen": -5.486642360687256, "rewards/margins": 1.6755880117416382, "rewards/rejected": -7.162230491638184, "step": 290 }, { "epoch": 0.644808743169399, "grad_norm": 66.70134041294685, "learning_rate": 3.3679439308082774e-07, "logits/chosen": -1.1284725666046143, "logits/rejected": -1.126910924911499, "logps/chosen": -0.5425226092338562, "logps/rejected": -0.7452016472816467, "loss": 1.7199, "rewards/accuracies": 0.8187500238418579, "rewards/chosen": -5.425226211547852, "rewards/margins": 2.026790142059326, "rewards/rejected": -7.452017307281494, "step": 295 }, { "epoch": 0.6557377049180327, "grad_norm": 71.88498470947849, "learning_rate": 3.1885519212446716e-07, "logits/chosen": -1.134820580482483, "logits/rejected": -1.113407850265503, "logps/chosen": -0.5564647316932678, "logps/rejected": -0.7468981146812439, "loss": 1.8137, "rewards/accuracies": 0.8187500238418579, "rewards/chosen": -5.564647197723389, "rewards/margins": 1.9043340682983398, "rewards/rejected": -7.4689812660217285, "step": 300 }, { "epoch": 0.6666666666666666, "grad_norm": 76.90950112576293, "learning_rate": 3.0118055418614295e-07, "logits/chosen": -1.1550134420394897, "logits/rejected": -1.0957601070404053, "logps/chosen": -0.5463107228279114, "logps/rejected": -0.7333046793937683, "loss": 1.942, "rewards/accuracies": 0.8125, "rewards/chosen": -5.463107109069824, "rewards/margins": 1.8699395656585693, "rewards/rejected": -7.333047389984131, "step": 305 }, { "epoch": 0.6775956284153005, "grad_norm": 107.13681953131237, "learning_rate": 2.83796293174686e-07, "logits/chosen": -1.0895421504974365, "logits/rejected": -1.0931763648986816, "logps/chosen": -0.5538777112960815, "logps/rejected": -0.7688311338424683, "loss": 2.0538, "rewards/accuracies": 0.7437499761581421, "rewards/chosen": -5.538776874542236, "rewards/margins": 2.149533748626709, "rewards/rejected": -7.688311576843262, "step": 310 }, { "epoch": 0.6885245901639344, "grad_norm": 66.34265535656564, "learning_rate": 2.6672779890178046e-07, "logits/chosen": -1.1496803760528564, "logits/rejected": -1.1489367485046387, "logps/chosen": -0.5764937400817871, "logps/rejected": -0.706534206867218, "loss": 1.9103, "rewards/accuracies": 0.737500011920929, "rewards/chosen": -5.764937400817871, "rewards/margins": 1.3004045486450195, "rewards/rejected": -7.065341949462891, "step": 315 }, { "epoch": 0.6994535519125683, "grad_norm": 56.65585571674297, "learning_rate": 2.500000000000001e-07, "logits/chosen": -1.204594612121582, "logits/rejected": -1.1554490327835083, "logps/chosen": -0.556768000125885, "logps/rejected": -0.7628398537635803, "loss": 1.8878, "rewards/accuracies": 0.824999988079071, "rewards/chosen": -5.567679405212402, "rewards/margins": 2.060718536376953, "rewards/rejected": -7.628398895263672, "step": 320 }, { "epoch": 0.7103825136612022, "grad_norm": 82.81929805474688, "learning_rate": 2.3363732751439923e-07, "logits/chosen": -1.1415125131607056, "logits/rejected": -1.125905990600586, "logps/chosen": -0.542820930480957, "logps/rejected": -0.7379633188247681, "loss": 1.8941, "rewards/accuracies": 0.793749988079071, "rewards/chosen": -5.42820930480957, "rewards/margins": 1.9514230489730835, "rewards/rejected": -7.379631996154785, "step": 325 }, { "epoch": 0.7213114754098361, "grad_norm": 71.70622872848305, "learning_rate": 2.1766367922083283e-07, "logits/chosen": -1.0906676054000854, "logits/rejected": -1.0594781637191772, "logps/chosen": -0.5101396441459656, "logps/rejected": -0.7594174146652222, "loss": 1.8536, "rewards/accuracies": 0.8062499761581421, "rewards/chosen": -5.101396083831787, "rewards/margins": 2.4927775859832764, "rewards/rejected": -7.594174385070801, "step": 330 }, { "epoch": 0.73224043715847, "grad_norm": 80.29511185962221, "learning_rate": 2.021023847231202e-07, "logits/chosen": -1.0921766757965088, "logits/rejected": -1.0546457767486572, "logps/chosen": -0.5786954760551453, "logps/rejected": -0.775272011756897, "loss": 1.859, "rewards/accuracies": 0.824999988079071, "rewards/chosen": -5.7869553565979, "rewards/margins": 1.9657647609710693, "rewards/rejected": -7.752719879150391, "step": 335 }, { "epoch": 0.7431693989071039, "grad_norm": 79.83892973093212, "learning_rate": 1.869761713800254e-07, "logits/chosen": -1.1134289503097534, "logits/rejected": -1.0689969062805176, "logps/chosen": -0.5616486668586731, "logps/rejected": -0.7319185137748718, "loss": 1.9172, "rewards/accuracies": 0.7250000238418579, "rewards/chosen": -5.6164870262146, "rewards/margins": 1.7026973962783813, "rewards/rejected": -7.319184303283691, "step": 340 }, { "epoch": 0.7540983606557377, "grad_norm": 79.67263653958703, "learning_rate": 1.7230713111182164e-07, "logits/chosen": -1.1623106002807617, "logits/rejected": -1.1600934267044067, "logps/chosen": -0.5594975352287292, "logps/rejected": -0.7684382200241089, "loss": 1.9557, "rewards/accuracies": 0.8125, "rewards/chosen": -5.594975471496582, "rewards/margins": 2.089407444000244, "rewards/rejected": -7.684383392333984, "step": 345 }, { "epoch": 0.7650273224043715, "grad_norm": 90.3235756484309, "learning_rate": 1.5811668813491696e-07, "logits/chosen": -1.1584365367889404, "logits/rejected": -1.1381675004959106, "logps/chosen": -0.5266227126121521, "logps/rejected": -0.6895478367805481, "loss": 1.8798, "rewards/accuracies": 0.800000011920929, "rewards/chosen": -5.266226768493652, "rewards/margins": 1.6292508840560913, "rewards/rejected": -6.895478248596191, "step": 350 }, { "epoch": 0.7759562841530054, "grad_norm": 65.15244365555583, "learning_rate": 1.4442556767166369e-07, "logits/chosen": -1.1325640678405762, "logits/rejected": -1.10225510597229, "logps/chosen": -0.539125382900238, "logps/rejected": -0.7123144865036011, "loss": 1.8691, "rewards/accuracies": 0.800000011920929, "rewards/chosen": -5.391254425048828, "rewards/margins": 1.731890082359314, "rewards/rejected": -7.123144626617432, "step": 355 }, { "epoch": 0.7868852459016393, "grad_norm": 74.4651421428626, "learning_rate": 1.312537656810549e-07, "logits/chosen": -1.0851835012435913, "logits/rejected": -1.0858454704284668, "logps/chosen": -0.5476486682891846, "logps/rejected": -0.730248749256134, "loss": 1.9346, "rewards/accuracies": 0.737500011920929, "rewards/chosen": -5.476487159729004, "rewards/margins": 1.8260005712509155, "rewards/rejected": -7.302487373352051, "step": 360 }, { "epoch": 0.7978142076502732, "grad_norm": 153.86057279690542, "learning_rate": 1.1862051965451214e-07, "logits/chosen": -1.166272521018982, "logits/rejected": -1.165235161781311, "logps/chosen": -0.5564864873886108, "logps/rejected": -0.749941349029541, "loss": 1.8628, "rewards/accuracies": 0.8187500238418579, "rewards/chosen": -5.5648651123046875, "rewards/margins": 1.934548020362854, "rewards/rejected": -7.49941349029541, "step": 365 }, { "epoch": 0.8087431693989071, "grad_norm": 77.62021036106599, "learning_rate": 1.0654428051942138e-07, "logits/chosen": -1.1702772378921509, "logits/rejected": -1.1325637102127075, "logps/chosen": -0.5678818225860596, "logps/rejected": -0.7951611876487732, "loss": 1.9704, "rewards/accuracies": 0.78125, "rewards/chosen": -5.678818225860596, "rewards/margins": 2.272794246673584, "rewards/rejected": -7.951612949371338, "step": 370 }, { "epoch": 0.819672131147541, "grad_norm": 63.25811531390013, "learning_rate": 9.504268569144763e-08, "logits/chosen": -1.2080334424972534, "logits/rejected": -1.1406437158584595, "logps/chosen": -0.5438567399978638, "logps/rejected": -0.7369149923324585, "loss": 1.892, "rewards/accuracies": 0.762499988079071, "rewards/chosen": -5.4385666847229, "rewards/margins": 1.9305833578109741, "rewards/rejected": -7.369150638580322, "step": 375 }, { "epoch": 0.8306010928961749, "grad_norm": 62.759245925109596, "learning_rate": 8.413253331499049e-08, "logits/chosen": -1.088577389717102, "logits/rejected": -1.1100099086761475, "logps/chosen": -0.56348717212677, "logps/rejected": -0.7516815066337585, "loss": 1.8378, "rewards/accuracies": 0.862500011920929, "rewards/chosen": -5.634871959686279, "rewards/margins": 1.8819434642791748, "rewards/rejected": -7.516815185546875, "step": 380 }, { "epoch": 0.8415300546448088, "grad_norm": 66.42055041642078, "learning_rate": 7.382975772939865e-08, "logits/chosen": -1.182722806930542, "logits/rejected": -1.1655415296554565, "logps/chosen": -0.599307656288147, "logps/rejected": -0.7887434959411621, "loss": 1.9674, "rewards/accuracies": 0.8500000238418579, "rewards/chosen": -5.993075847625732, "rewards/margins": 1.8943592309951782, "rewards/rejected": -7.887434959411621, "step": 385 }, { "epoch": 0.8524590163934426, "grad_norm": 104.21611140368309, "learning_rate": 6.414940619677734e-08, "logits/chosen": -1.171216607093811, "logits/rejected": -1.1473820209503174, "logps/chosen": -0.5480982065200806, "logps/rejected": -0.76609867811203, "loss": 1.9188, "rewards/accuracies": 0.831250011920929, "rewards/chosen": -5.480981349945068, "rewards/margins": 2.1800050735473633, "rewards/rejected": -7.660986423492432, "step": 390 }, { "epoch": 0.8633879781420765, "grad_norm": 78.27102698318478, "learning_rate": 5.5105616925376296e-08, "logits/chosen": -1.1506216526031494, "logits/rejected": -1.1313217878341675, "logps/chosen": -0.5556925535202026, "logps/rejected": -0.706631064414978, "loss": 1.8449, "rewards/accuracies": 0.768750011920929, "rewards/chosen": -5.556925296783447, "rewards/margins": 1.509385347366333, "rewards/rejected": -7.066309928894043, "step": 395 }, { "epoch": 0.8743169398907104, "grad_norm": 76.30561413947063, "learning_rate": 4.6711598420656976e-08, "logits/chosen": -1.0807750225067139, "logits/rejected": -1.0536285638809204, "logps/chosen": -0.5846288800239563, "logps/rejected": -0.7902191281318665, "loss": 1.8354, "rewards/accuracies": 0.8062499761581421, "rewards/chosen": -5.846289157867432, "rewards/margins": 2.0559017658233643, "rewards/rejected": -7.902191162109375, "step": 400 }, { "epoch": 0.8743169398907104, "eval_logits/chosen": -1.3447647094726562, "eval_logits/rejected": -1.3069871664047241, "eval_logps/chosen": -0.5437440872192383, "eval_logps/rejected": -0.7271364331245422, "eval_loss": 1.881577491760254, "eval_rewards/accuracies": 0.7710843086242676, "eval_rewards/chosen": -5.437441349029541, "eval_rewards/margins": 1.83392333984375, "eval_rewards/rejected": -7.271363735198975, "eval_runtime": 37.6564, "eval_samples_per_second": 35.001, "eval_steps_per_second": 2.204, "step": 400 }, { "epoch": 0.8852459016393442, "grad_norm": 66.16858739679557, "learning_rate": 3.897961019419516e-08, "logits/chosen": -1.1171514987945557, "logits/rejected": -1.051194190979004, "logps/chosen": -0.5250142812728882, "logps/rejected": -0.6711951494216919, "loss": 1.793, "rewards/accuracies": 0.7749999761581421, "rewards/chosen": -5.250143527984619, "rewards/margins": 1.461808204650879, "rewards/rejected": -6.71195125579834, "step": 405 }, { "epoch": 0.8961748633879781, "grad_norm": 77.0310414767933, "learning_rate": 3.192094485859526e-08, "logits/chosen": -1.1234816312789917, "logits/rejected": -1.144008994102478, "logps/chosen": -0.5635503530502319, "logps/rejected": -0.7934846878051758, "loss": 1.8728, "rewards/accuracies": 0.7749999761581421, "rewards/chosen": -5.63550329208374, "rewards/margins": 2.2993433475494385, "rewards/rejected": -7.934847354888916, "step": 410 }, { "epoch": 0.907103825136612, "grad_norm": 62.5404158274799, "learning_rate": 2.5545911634565265e-08, "logits/chosen": -1.1487548351287842, "logits/rejected": -1.1474642753601074, "logps/chosen": -0.570233941078186, "logps/rejected": -0.7942467927932739, "loss": 1.8971, "rewards/accuracies": 0.824999988079071, "rewards/chosen": -5.702339172363281, "rewards/margins": 2.2401280403137207, "rewards/rejected": -7.94246768951416, "step": 415 }, { "epoch": 0.9180327868852459, "grad_norm": 81.2075081960254, "learning_rate": 1.9863821294241522e-08, "logits/chosen": -1.150827407836914, "logits/rejected": -1.1209652423858643, "logps/chosen": -0.5287235379219055, "logps/rejected": -0.7498638033866882, "loss": 1.7791, "rewards/accuracies": 0.8125, "rewards/chosen": -5.287235260009766, "rewards/margins": 2.211402177810669, "rewards/rejected": -7.498638153076172, "step": 420 }, { "epoch": 0.9289617486338798, "grad_norm": 74.25861489169881, "learning_rate": 1.4882972562753615e-08, "logits/chosen": -1.1429412364959717, "logits/rejected": -1.1335225105285645, "logps/chosen": -0.5883482098579407, "logps/rejected": -0.8031044006347656, "loss": 1.918, "rewards/accuracies": 0.7875000238418579, "rewards/chosen": -5.883482933044434, "rewards/margins": 2.1475605964660645, "rewards/rejected": -8.03104305267334, "step": 425 }, { "epoch": 0.9398907103825137, "grad_norm": 90.08434920268098, "learning_rate": 1.0610639997888915e-08, "logits/chosen": -1.0862069129943848, "logits/rejected": -1.086482286453247, "logps/chosen": -0.5331538915634155, "logps/rejected": -0.73698890209198, "loss": 1.7152, "rewards/accuracies": 0.8812500238418579, "rewards/chosen": -5.331539154052734, "rewards/margins": 2.0383496284484863, "rewards/rejected": -7.3698883056640625, "step": 430 }, { "epoch": 0.9508196721311475, "grad_norm": 88.78145235758741, "learning_rate": 7.053063365559997e-09, "logits/chosen": -1.1540377140045166, "logits/rejected": -1.1776275634765625, "logps/chosen": -0.5493609309196472, "logps/rejected": -0.7795323729515076, "loss": 1.8391, "rewards/accuracies": 0.8500000238418579, "rewards/chosen": -5.493609428405762, "rewards/margins": 2.3017146587371826, "rewards/rejected": -7.795323848724365, "step": 435 }, { "epoch": 0.9617486338797814, "grad_norm": 76.77833268976481, "learning_rate": 4.215438526591064e-09, "logits/chosen": -1.091371774673462, "logits/rejected": -1.0515260696411133, "logps/chosen": -0.5720237493515015, "logps/rejected": -0.7283024787902832, "loss": 1.8312, "rewards/accuracies": 0.762499988079071, "rewards/chosen": -5.720236778259277, "rewards/margins": 1.5627872943878174, "rewards/rejected": -7.283023834228516, "step": 440 }, { "epoch": 0.9726775956284153, "grad_norm": 85.67571762449911, "learning_rate": 2.1019098481337426e-09, "logits/chosen": -1.1658837795257568, "logits/rejected": -1.1402862071990967, "logps/chosen": -0.5455365180969238, "logps/rejected": -0.7272461652755737, "loss": 1.8014, "rewards/accuracies": 0.8187500238418579, "rewards/chosen": -5.4553656578063965, "rewards/margins": 1.8170963525772095, "rewards/rejected": -7.272461891174316, "step": 445 }, { "epoch": 0.9836065573770492, "grad_norm": 87.49876973882085, "learning_rate": 7.155641507955445e-10, "logits/chosen": -1.0767357349395752, "logits/rejected": -1.0696969032287598, "logps/chosen": -0.5918444395065308, "logps/rejected": -0.7698336839675903, "loss": 1.9461, "rewards/accuracies": 0.78125, "rewards/chosen": -5.918444633483887, "rewards/margins": 1.7798923254013062, "rewards/rejected": -7.698336601257324, "step": 450 }, { "epoch": 0.994535519125683, "grad_norm": 83.15276885308347, "learning_rate": 5.842620032053824e-11, "logits/chosen": -1.0845460891723633, "logits/rejected": -1.0800822973251343, "logps/chosen": -0.5885196328163147, "logps/rejected": -0.7454328536987305, "loss": 1.9848, "rewards/accuracies": 0.7562500238418579, "rewards/chosen": -5.885196208953857, "rewards/margins": 1.5691322088241577, "rewards/rejected": -7.4543280601501465, "step": 455 }, { "epoch": 0.9989071038251366, "step": 457, "total_flos": 0.0, "train_loss": 2.2740620700744363, "train_runtime": 5957.0658, "train_samples_per_second": 9.83, "train_steps_per_second": 0.077 } ], "logging_steps": 5, "max_steps": 457, "num_input_tokens_seen": 0, "num_train_epochs": 1, "save_steps": 1000000, "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": 2, "trial_name": null, "trial_params": null }