{ "best_metric": null, "best_model_checkpoint": null, "epoch": 2.9865871833084947, "eval_steps": 500, "global_step": 501, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.029806259314456036, "grad_norm": 1802.2630615234375, "learning_rate": 3.75e-07, "log_odds_chosen": -0.22044436633586884, "log_odds_ratio": -1.0070970058441162, "logits/chosen": 204.41079711914062, "logits/rejected": 203.1068878173828, "logps/chosen": -14.805323600769043, "logps/rejected": -14.584878921508789, "loss": 15.3769, "nll_loss": 14.526931762695312, "rewards/accuracies": 0.375, "rewards/chosen": -7.4026618003845215, "rewards/margins": -0.11022226512432098, "rewards/rejected": -7.2924394607543945, "step": 5 }, { "epoch": 0.05961251862891207, "grad_norm": 964.4558715820312, "learning_rate": 7.5e-07, "log_odds_chosen": 0.22153806686401367, "log_odds_ratio": -0.7704828977584839, "logits/chosen": 229.1735382080078, "logits/rejected": 234.7494659423828, "logps/chosen": -10.87135124206543, "logps/rejected": -11.092044830322266, "loss": 11.644, "nll_loss": 10.984331130981445, "rewards/accuracies": 0.5, "rewards/chosen": -5.435675621032715, "rewards/margins": 0.1103477030992508, "rewards/rejected": -5.546022415161133, "step": 10 }, { "epoch": 0.08941877794336811, "grad_norm": 325.32958984375, "learning_rate": 1.125e-06, "log_odds_chosen": 0.0398867093026638, "log_odds_ratio": -0.7268143892288208, "logits/chosen": 288.3202209472656, "logits/rejected": 268.65325927734375, "logps/chosen": -6.712647914886475, "logps/rejected": -6.752583980560303, "loss": 7.2706, "nll_loss": 6.687516689300537, "rewards/accuracies": 0.550000011920929, "rewards/chosen": -3.3563239574432373, "rewards/margins": 0.019967829808592796, "rewards/rejected": -3.3762919902801514, "step": 15 }, { "epoch": 0.11922503725782414, "grad_norm": 153.8790740966797, "learning_rate": 1.5e-06, "log_odds_chosen": 0.005181239452213049, "log_odds_ratio": -0.8698928952217102, "logits/chosen": 279.0413513183594, "logits/rejected": 273.7066955566406, "logps/chosen": -4.006048202514648, "logps/rejected": -4.012642860412598, "loss": 4.5269, "nll_loss": 4.109046459197998, "rewards/accuracies": 0.44999998807907104, "rewards/chosen": -2.003024101257324, "rewards/margins": 0.003297415329143405, "rewards/rejected": -2.006321430206299, "step": 20 }, { "epoch": 0.14903129657228018, "grad_norm": 110.71721649169922, "learning_rate": 1.875e-06, "log_odds_chosen": -0.0372297465801239, "log_odds_ratio": -0.8627766370773315, "logits/chosen": 340.18084716796875, "logits/rejected": 349.11151123046875, "logps/chosen": -2.547694206237793, "logps/rejected": -2.4806361198425293, "loss": 3.1921, "nll_loss": 2.749494791030884, "rewards/accuracies": 0.5, "rewards/chosen": -1.2738471031188965, "rewards/margins": -0.03352922201156616, "rewards/rejected": -1.2403180599212646, "step": 25 }, { "epoch": 0.17883755588673622, "grad_norm": 75.32969665527344, "learning_rate": 2.25e-06, "log_odds_chosen": 0.018188167363405228, "log_odds_ratio": -0.8359094858169556, "logits/chosen": 373.6573791503906, "logits/rejected": 400.1236572265625, "logps/chosen": -2.330045223236084, "logps/rejected": -2.3124964237213135, "loss": 2.6112, "nll_loss": 2.3758959770202637, "rewards/accuracies": 0.5249999761581421, "rewards/chosen": -1.165022611618042, "rewards/margins": -0.008774325251579285, "rewards/rejected": -1.1562482118606567, "step": 30 }, { "epoch": 0.20864381520119224, "grad_norm": 45.19425964355469, "learning_rate": 2.625e-06, "log_odds_chosen": 0.25173941254615784, "log_odds_ratio": -0.7035487294197083, "logits/chosen": 391.4384765625, "logits/rejected": 380.1146545410156, "logps/chosen": -1.7061666250228882, "logps/rejected": -1.8836677074432373, "loss": 2.3484, "nll_loss": 1.9240200519561768, "rewards/accuracies": 0.574999988079071, "rewards/chosen": -0.8530833125114441, "rewards/margins": 0.08875055611133575, "rewards/rejected": -0.9418338537216187, "step": 35 }, { "epoch": 0.23845007451564829, "grad_norm": 49.53385925292969, "learning_rate": 3e-06, "log_odds_chosen": 0.2747120261192322, "log_odds_ratio": -0.6599891185760498, "logits/chosen": 381.4653625488281, "logits/rejected": 379.7186279296875, "logps/chosen": -1.598753571510315, "logps/rejected": -1.8254070281982422, "loss": 2.2488, "nll_loss": 2.010526418685913, "rewards/accuracies": 0.5249999761581421, "rewards/chosen": -0.7993767857551575, "rewards/margins": 0.11332666873931885, "rewards/rejected": -0.9127035140991211, "step": 40 }, { "epoch": 0.26825633383010433, "grad_norm": 68.84819030761719, "learning_rate": 3.3750000000000003e-06, "log_odds_chosen": 0.29411372542381287, "log_odds_ratio": -0.6252076625823975, "logits/chosen": 395.29290771484375, "logits/rejected": 404.8909912109375, "logps/chosen": -1.651572823524475, "logps/rejected": -1.8901698589324951, "loss": 2.1795, "nll_loss": 1.7518342733383179, "rewards/accuracies": 0.6499999761581421, "rewards/chosen": -0.8257864117622375, "rewards/margins": 0.11929843574762344, "rewards/rejected": -0.9450849294662476, "step": 45 }, { "epoch": 0.29806259314456035, "grad_norm": 47.519283294677734, "learning_rate": 3.75e-06, "log_odds_chosen": 0.5501509308815002, "log_odds_ratio": -0.5450042486190796, "logits/chosen": 403.5310363769531, "logits/rejected": 424.2149963378906, "logps/chosen": -1.5113065242767334, "logps/rejected": -1.964956521987915, "loss": 2.1594, "nll_loss": 1.8876510858535767, "rewards/accuracies": 0.699999988079071, "rewards/chosen": -0.7556532621383667, "rewards/margins": 0.22682484984397888, "rewards/rejected": -0.9824782609939575, "step": 50 }, { "epoch": 0.32786885245901637, "grad_norm": 36.98734664916992, "learning_rate": 4.125e-06, "log_odds_chosen": 0.26053518056869507, "log_odds_ratio": -0.6607739925384521, "logits/chosen": 388.755859375, "logits/rejected": 381.29827880859375, "logps/chosen": -1.498511552810669, "logps/rejected": -1.7075812816619873, "loss": 2.1979, "nll_loss": 1.9259532690048218, "rewards/accuracies": 0.550000011920929, "rewards/chosen": -0.7492557764053345, "rewards/margins": 0.1045350432395935, "rewards/rejected": -0.8537906408309937, "step": 55 }, { "epoch": 0.35767511177347244, "grad_norm": 63.64994430541992, "learning_rate": 4.5e-06, "log_odds_chosen": 0.23358798027038574, "log_odds_ratio": -0.6189488172531128, "logits/chosen": 396.7557678222656, "logits/rejected": 388.23681640625, "logps/chosen": -1.4873348474502563, "logps/rejected": -1.6665433645248413, "loss": 2.0244, "nll_loss": 1.826873540878296, "rewards/accuracies": 0.625, "rewards/chosen": -0.7436674237251282, "rewards/margins": 0.08960425853729248, "rewards/rejected": -0.8332716822624207, "step": 60 }, { "epoch": 0.38748137108792846, "grad_norm": 27.614044189453125, "learning_rate": 4.875e-06, "log_odds_chosen": 0.3461737036705017, "log_odds_ratio": -0.6412385106086731, "logits/chosen": 397.4202575683594, "logits/rejected": 397.5688171386719, "logps/chosen": -1.3429419994354248, "logps/rejected": -1.6044105291366577, "loss": 1.945, "nll_loss": 1.5645616054534912, "rewards/accuracies": 0.6499999761581421, "rewards/chosen": -0.6714709997177124, "rewards/margins": 0.13073426485061646, "rewards/rejected": -0.8022052645683289, "step": 65 }, { "epoch": 0.4172876304023845, "grad_norm": 86.65094757080078, "learning_rate": 5.25e-06, "log_odds_chosen": 0.41045960783958435, "log_odds_ratio": -0.5924609303474426, "logits/chosen": 376.8070983886719, "logits/rejected": 384.87994384765625, "logps/chosen": -1.5336923599243164, "logps/rejected": -1.8701730966567993, "loss": 1.997, "nll_loss": 1.7713584899902344, "rewards/accuracies": 0.699999988079071, "rewards/chosen": -0.7668461799621582, "rewards/margins": 0.16824033856391907, "rewards/rejected": -0.9350865483283997, "step": 70 }, { "epoch": 0.44709388971684055, "grad_norm": 75.32307434082031, "learning_rate": 5.625e-06, "log_odds_chosen": 0.5093179941177368, "log_odds_ratio": -0.5533502101898193, "logits/chosen": 403.20697021484375, "logits/rejected": 389.50067138671875, "logps/chosen": -1.303440809249878, "logps/rejected": -1.7079664468765259, "loss": 1.901, "nll_loss": 1.5844974517822266, "rewards/accuracies": 0.75, "rewards/chosen": -0.651720404624939, "rewards/margins": 0.20226280391216278, "rewards/rejected": -0.8539832234382629, "step": 75 }, { "epoch": 0.47690014903129657, "grad_norm": 52.309078216552734, "learning_rate": 6e-06, "log_odds_chosen": 0.5146449208259583, "log_odds_ratio": -0.5773543119430542, "logits/chosen": 392.9078063964844, "logits/rejected": 413.81451416015625, "logps/chosen": -1.4215633869171143, "logps/rejected": -1.8292858600616455, "loss": 1.9712, "nll_loss": 1.699523687362671, "rewards/accuracies": 0.6499999761581421, "rewards/chosen": -0.7107816934585571, "rewards/margins": 0.20386135578155518, "rewards/rejected": -0.9146429300308228, "step": 80 }, { "epoch": 0.5067064083457526, "grad_norm": 665.8851318359375, "learning_rate": 6.375e-06, "log_odds_chosen": 0.4371088147163391, "log_odds_ratio": -0.5891958475112915, "logits/chosen": 411.912109375, "logits/rejected": 395.8421325683594, "logps/chosen": -1.437474012374878, "logps/rejected": -1.776432991027832, "loss": 2.0743, "nll_loss": 1.8243602514266968, "rewards/accuracies": 0.675000011920929, "rewards/chosen": -0.718737006187439, "rewards/margins": 0.16947945952415466, "rewards/rejected": -0.888216495513916, "step": 85 }, { "epoch": 0.5365126676602087, "grad_norm": 57.34864044189453, "learning_rate": 6.750000000000001e-06, "log_odds_chosen": 0.9314719438552856, "log_odds_ratio": -0.49316421151161194, "logits/chosen": 410.67156982421875, "logits/rejected": 423.86431884765625, "logps/chosen": -1.415283441543579, "logps/rejected": -2.1992759704589844, "loss": 1.8288, "nll_loss": 1.5969065427780151, "rewards/accuracies": 0.7250000238418579, "rewards/chosen": -0.7076417207717896, "rewards/margins": 0.39199626445770264, "rewards/rejected": -1.0996379852294922, "step": 90 }, { "epoch": 0.5663189269746647, "grad_norm": 65.98050689697266, "learning_rate": 7.1249999999999995e-06, "log_odds_chosen": 0.38506197929382324, "log_odds_ratio": -0.5869079828262329, "logits/chosen": 378.6953430175781, "logits/rejected": 390.01422119140625, "logps/chosen": -1.212702989578247, "logps/rejected": -1.4467040300369263, "loss": 1.8818, "nll_loss": 1.4865398406982422, "rewards/accuracies": 0.6000000238418579, "rewards/chosen": -0.6063514947891235, "rewards/margins": 0.1170005351305008, "rewards/rejected": -0.7233520150184631, "step": 95 }, { "epoch": 0.5961251862891207, "grad_norm": 42.52804183959961, "learning_rate": 7.5e-06, "log_odds_chosen": 0.6413298845291138, "log_odds_ratio": -0.538486659526825, "logits/chosen": 412.3399353027344, "logits/rejected": 447.61865234375, "logps/chosen": -1.3371002674102783, "logps/rejected": -1.868131399154663, "loss": 1.9035, "nll_loss": 1.6383569240570068, "rewards/accuracies": 0.7250000238418579, "rewards/chosen": -0.6685501337051392, "rewards/margins": 0.2655157446861267, "rewards/rejected": -0.9340656995773315, "step": 100 }, { "epoch": 0.6259314456035767, "grad_norm": 48.187232971191406, "learning_rate": 7.319250547113999e-06, "log_odds_chosen": 0.7124954462051392, "log_odds_ratio": -0.5506631731987, "logits/chosen": 403.2848205566406, "logits/rejected": 415.0179138183594, "logps/chosen": -1.253138780593872, "logps/rejected": -1.768273115158081, "loss": 1.8545, "nll_loss": 1.6297533512115479, "rewards/accuracies": 0.699999988079071, "rewards/chosen": -0.626569390296936, "rewards/margins": 0.25756722688674927, "rewards/rejected": -0.8841365575790405, "step": 105 }, { "epoch": 0.6557377049180327, "grad_norm": 195.05877685546875, "learning_rate": 7.150969419341942e-06, "log_odds_chosen": 0.4548686444759369, "log_odds_ratio": -0.569240927696228, "logits/chosen": 389.09381103515625, "logits/rejected": 386.32513427734375, "logps/chosen": -1.3020490407943726, "logps/rejected": -1.6546863317489624, "loss": 1.8713, "nll_loss": 1.6556203365325928, "rewards/accuracies": 0.675000011920929, "rewards/chosen": -0.6510245203971863, "rewards/margins": 0.1763187050819397, "rewards/rejected": -0.8273431658744812, "step": 110 }, { "epoch": 0.6855439642324889, "grad_norm": 53.028961181640625, "learning_rate": 6.993786061802354e-06, "log_odds_chosen": 0.8375183939933777, "log_odds_ratio": -0.4603613018989563, "logits/chosen": 408.5062561035156, "logits/rejected": 414.22332763671875, "logps/chosen": -1.220641851425171, "logps/rejected": -1.8461415767669678, "loss": 1.8867, "nll_loss": 1.583599328994751, "rewards/accuracies": 0.7749999761581421, "rewards/chosen": -0.6103209257125854, "rewards/margins": 0.3127499222755432, "rewards/rejected": -0.9230707883834839, "step": 115 }, { "epoch": 0.7153502235469449, "grad_norm": 49.91960144042969, "learning_rate": 6.846531968814577e-06, "log_odds_chosen": 0.6342799663543701, "log_odds_ratio": -0.5294886827468872, "logits/chosen": 393.26812744140625, "logits/rejected": 391.69677734375, "logps/chosen": -1.1840012073516846, "logps/rejected": -1.6264234781265259, "loss": 1.9171, "nll_loss": 1.6122639179229736, "rewards/accuracies": 0.7250000238418579, "rewards/chosen": -0.5920006036758423, "rewards/margins": 0.22121115028858185, "rewards/rejected": -0.8132117390632629, "step": 120 }, { "epoch": 0.7451564828614009, "grad_norm": 26.490934371948242, "learning_rate": 6.708203932499369e-06, "log_odds_chosen": 0.3495193421840668, "log_odds_ratio": -0.6338833570480347, "logits/chosen": 394.9533386230469, "logits/rejected": 406.28662109375, "logps/chosen": -1.23225998878479, "logps/rejected": -1.4777755737304688, "loss": 1.8239, "nll_loss": 1.5472067594528198, "rewards/accuracies": 0.6000000238418579, "rewards/chosen": -0.616129994392395, "rewards/margins": 0.12275775521993637, "rewards/rejected": -0.7388877868652344, "step": 125 }, { "epoch": 0.7749627421758569, "grad_norm": 45.540184020996094, "learning_rate": 6.577935144802719e-06, "log_odds_chosen": 0.4649319052696228, "log_odds_ratio": -0.5916207432746887, "logits/chosen": 417.3863830566406, "logits/rejected": 405.1355895996094, "logps/chosen": -1.2279858589172363, "logps/rejected": -1.5856168270111084, "loss": 1.8861, "nll_loss": 1.565514326095581, "rewards/accuracies": 0.5249999761581421, "rewards/chosen": -0.6139929294586182, "rewards/margins": 0.17881537973880768, "rewards/rejected": -0.7928084135055542, "step": 130 }, { "epoch": 0.8047690014903129, "grad_norm": 42.16865921020508, "learning_rate": 6.454972243679028e-06, "log_odds_chosen": 0.46650758385658264, "log_odds_ratio": -0.6158651113510132, "logits/chosen": 421.04730224609375, "logits/rejected": 423.00775146484375, "logps/chosen": -1.509341835975647, "logps/rejected": -1.8879953622817993, "loss": 1.9281, "nll_loss": 1.723570466041565, "rewards/accuracies": 0.6499999761581421, "rewards/chosen": -0.7546709179878235, "rewards/margins": 0.18932664394378662, "rewards/rejected": -0.9439976811408997, "step": 135 }, { "epoch": 0.834575260804769, "grad_norm": 41.30878448486328, "learning_rate": 6.338656910463875e-06, "log_odds_chosen": 0.27211299538612366, "log_odds_ratio": -0.680214524269104, "logits/chosen": 404.1061096191406, "logits/rejected": 409.54559326171875, "logps/chosen": -1.4520971775054932, "logps/rejected": -1.670898199081421, "loss": 1.9311, "nll_loss": 1.7876545190811157, "rewards/accuracies": 0.5249999761581421, "rewards/chosen": -0.7260485887527466, "rewards/margins": 0.10940053313970566, "rewards/rejected": -0.8354490995407104, "step": 140 }, { "epoch": 0.8643815201192251, "grad_norm": 58.79216003417969, "learning_rate": 6.228410989030498e-06, "log_odds_chosen": 0.13751578330993652, "log_odds_ratio": -0.6719902753829956, "logits/chosen": 402.82342529296875, "logits/rejected": 405.1291809082031, "logps/chosen": -1.3680763244628906, "logps/rejected": -1.4593302011489868, "loss": 1.9064, "nll_loss": 1.6308927536010742, "rewards/accuracies": 0.6499999761581421, "rewards/chosen": -0.6840381622314453, "rewards/margins": 0.045626964420080185, "rewards/rejected": -0.7296651005744934, "step": 145 }, { "epoch": 0.8941877794336811, "grad_norm": 28.582611083984375, "learning_rate": 6.123724356957946e-06, "log_odds_chosen": 0.8705786466598511, "log_odds_ratio": -0.5025649666786194, "logits/chosen": 407.1871643066406, "logits/rejected": 423.8992614746094, "logps/chosen": -1.2447984218597412, "logps/rejected": -1.9607397317886353, "loss": 1.8397, "nll_loss": 1.4880900382995605, "rewards/accuracies": 0.75, "rewards/chosen": -0.6223992109298706, "rewards/margins": 0.357970654964447, "rewards/rejected": -0.9803698658943176, "step": 150 }, { "epoch": 0.9239940387481371, "grad_norm": 56.87172317504883, "learning_rate": 6.024144966768742e-06, "log_odds_chosen": 0.32878491282463074, "log_odds_ratio": -0.6313835382461548, "logits/chosen": 388.4563293457031, "logits/rejected": 402.23834228515625, "logps/chosen": -1.365476131439209, "logps/rejected": -1.6416184902191162, "loss": 1.8579, "nll_loss": 1.458609938621521, "rewards/accuracies": 0.6499999761581421, "rewards/chosen": -0.6827380657196045, "rewards/margins": 0.13807126879692078, "rewards/rejected": -0.8208092451095581, "step": 155 }, { "epoch": 0.9538002980625931, "grad_norm": 52.100013732910156, "learning_rate": 5.929270612815711e-06, "log_odds_chosen": 0.2496444433927536, "log_odds_ratio": -0.6904889345169067, "logits/chosen": 397.15887451171875, "logits/rejected": 440.45587158203125, "logps/chosen": -1.2235276699066162, "logps/rejected": -1.4282524585723877, "loss": 1.8144, "nll_loss": 1.502598762512207, "rewards/accuracies": 0.574999988079071, "rewards/chosen": -0.6117638349533081, "rewards/margins": 0.10236239433288574, "rewards/rejected": -0.7141262292861938, "step": 160 }, { "epoch": 0.9836065573770492, "grad_norm": 62.34568786621094, "learning_rate": 5.838742081211422e-06, "log_odds_chosen": 0.25045713782310486, "log_odds_ratio": -0.6064125895500183, "logits/chosen": 407.72943115234375, "logits/rejected": 420.569580078125, "logps/chosen": -1.2549211978912354, "logps/rejected": -1.4284086227416992, "loss": 1.8398, "nll_loss": 1.5022889375686646, "rewards/accuracies": 0.625, "rewards/chosen": -0.6274605989456177, "rewards/margins": 0.0867437794804573, "rewards/rejected": -0.7142043113708496, "step": 165 }, { "epoch": 0.9955290611028316, "eval_log_odds_chosen": 0.25424009561538696, "eval_log_odds_ratio": -0.6749479174613953, "eval_logits/chosen": 331.1551513671875, "eval_logits/rejected": 302.9094543457031, "eval_logps/chosen": -1.1524564027786255, "eval_logps/rejected": -1.343080997467041, "eval_loss": 1.8982083797454834, "eval_nll_loss": 1.5454045534133911, "eval_rewards/accuracies": 0.5179855823516846, "eval_rewards/chosen": -0.5762282013893127, "eval_rewards/margins": 0.09531236439943314, "eval_rewards/rejected": -0.6715404987335205, "eval_runtime": 378.1633, "eval_samples_per_second": 1.462, "eval_steps_per_second": 0.368, "step": 167 }, { "epoch": 1.0134128166915053, "grad_norm": 19.05902862548828, "learning_rate": 5.752237416355279e-06, "log_odds_chosen": 0.8472374081611633, "log_odds_ratio": -0.5489901304244995, "logits/chosen": 389.7673645019531, "logits/rejected": 412.4244079589844, "logps/chosen": -1.1339842081069946, "logps/rejected": -1.749389410018921, "loss": 1.6458, "nll_loss": 1.2897634506225586, "rewards/accuracies": 0.7250000238418579, "rewards/chosen": -0.5669921040534973, "rewards/margins": 0.3077026307582855, "rewards/rejected": -0.8746947050094604, "step": 170 }, { "epoch": 1.0432190760059612, "grad_norm": 39.458702087402344, "learning_rate": 5.669467095138408e-06, "log_odds_chosen": 1.5121970176696777, "log_odds_ratio": -0.30622434616088867, "logits/chosen": 365.8326416015625, "logits/rejected": 406.28411865234375, "logps/chosen": -0.7569233775138855, "logps/rejected": -1.806539535522461, "loss": 1.2547, "nll_loss": 1.183380365371704, "rewards/accuracies": 0.8999999761581421, "rewards/chosen": -0.37846168875694275, "rewards/margins": 0.5248081088066101, "rewards/rejected": -0.9032697677612305, "step": 175 }, { "epoch": 1.0730253353204173, "grad_norm": 31.42799949645996, "learning_rate": 5.5901699437494744e-06, "log_odds_chosen": 1.5321556329727173, "log_odds_ratio": -0.2848857045173645, "logits/chosen": 364.0169677734375, "logits/rejected": 339.5393371582031, "logps/chosen": -0.8578670620918274, "logps/rejected": -1.9486223459243774, "loss": 1.3431, "nll_loss": 1.3115805387496948, "rewards/accuracies": 0.925000011920929, "rewards/chosen": -0.4289335310459137, "rewards/margins": 0.5453776717185974, "rewards/rejected": -0.9743111729621887, "step": 180 }, { "epoch": 1.1028315946348732, "grad_norm": 35.6085090637207, "learning_rate": 5.514109665703559e-06, "log_odds_chosen": 2.223374843597412, "log_odds_ratio": -0.2399306744337082, "logits/chosen": 396.5127868652344, "logits/rejected": 386.45831298828125, "logps/chosen": -0.8754903078079224, "logps/rejected": -2.4870636463165283, "loss": 1.2172, "nll_loss": 1.070143222808838, "rewards/accuracies": 0.949999988079071, "rewards/chosen": -0.4377451539039612, "rewards/margins": 0.8057867288589478, "rewards/rejected": -1.2435318231582642, "step": 185 }, { "epoch": 1.1326378539493294, "grad_norm": 22.422922134399414, "learning_rate": 5.441071875825088e-06, "log_odds_chosen": 1.2998098134994507, "log_odds_ratio": -0.32330313324928284, "logits/chosen": 364.8489074707031, "logits/rejected": 408.52972412109375, "logps/chosen": -0.870254397392273, "logps/rejected": -1.7498779296875, "loss": 1.2807, "nll_loss": 1.1046603918075562, "rewards/accuracies": 0.875, "rewards/chosen": -0.4351271986961365, "rewards/margins": 0.4398117661476135, "rewards/rejected": -0.87493896484375, "step": 190 }, { "epoch": 1.1624441132637853, "grad_norm": 25.98566246032715, "learning_rate": 5.370861555295747e-06, "log_odds_chosen": 2.002765417098999, "log_odds_ratio": -0.24665918946266174, "logits/chosen": 384.2296142578125, "logits/rejected": 398.67205810546875, "logps/chosen": -0.8954353332519531, "logps/rejected": -2.385274648666382, "loss": 1.2781, "nll_loss": 1.1818829774856567, "rewards/accuracies": 0.9750000238418579, "rewards/chosen": -0.44771766662597656, "rewards/margins": 0.7449195981025696, "rewards/rejected": -1.192637324333191, "step": 195 }, { "epoch": 1.1922503725782414, "grad_norm": 23.273193359375, "learning_rate": 5.303300858899106e-06, "log_odds_chosen": 2.1724612712860107, "log_odds_ratio": -0.26438647508621216, "logits/chosen": 413.26361083984375, "logits/rejected": 396.8680419921875, "logps/chosen": -0.730914831161499, "logps/rejected": -2.3943963050842285, "loss": 1.2803, "nll_loss": 1.0805118083953857, "rewards/accuracies": 0.949999988079071, "rewards/chosen": -0.3654574155807495, "rewards/margins": 0.83174067735672, "rewards/rejected": -1.1971981525421143, "step": 200 }, { "epoch": 1.2220566318926975, "grad_norm": 27.806596755981445, "learning_rate": 5.238227218271837e-06, "log_odds_chosen": 2.4484591484069824, "log_odds_ratio": -0.22746221721172333, "logits/chosen": 365.72119140625, "logits/rejected": 359.6267395019531, "logps/chosen": -0.758979320526123, "logps/rejected": -2.573237180709839, "loss": 1.2398, "nll_loss": 1.1106595993041992, "rewards/accuracies": 0.9750000238418579, "rewards/chosen": -0.3794896602630615, "rewards/margins": 0.9071289896965027, "rewards/rejected": -1.2866185903549194, "step": 205 }, { "epoch": 1.2518628912071534, "grad_norm": 22.43109130859375, "learning_rate": 5.175491695067656e-06, "log_odds_chosen": 2.793255567550659, "log_odds_ratio": -0.23325839638710022, "logits/chosen": 377.73443603515625, "logits/rejected": 404.66644287109375, "logps/chosen": -0.7964418530464172, "logps/rejected": -3.0091164112091064, "loss": 1.2371, "nll_loss": 1.0737245082855225, "rewards/accuracies": 0.949999988079071, "rewards/chosen": -0.3982209265232086, "rewards/margins": 1.106337308883667, "rewards/rejected": -1.5045582056045532, "step": 210 }, { "epoch": 1.2816691505216096, "grad_norm": 49.615081787109375, "learning_rate": 5.114957546028551e-06, "log_odds_chosen": 2.5913193225860596, "log_odds_ratio": -0.23048730194568634, "logits/chosen": 401.5857238769531, "logits/rejected": 382.7472229003906, "logps/chosen": -0.7830225229263306, "logps/rejected": -2.782843828201294, "loss": 1.2555, "nll_loss": 1.1490776538848877, "rewards/accuracies": 0.949999988079071, "rewards/chosen": -0.3915112614631653, "rewards/margins": 0.9999106526374817, "rewards/rejected": -1.391421914100647, "step": 215 }, { "epoch": 1.3114754098360657, "grad_norm": 37.056453704833984, "learning_rate": 5.056498968474315e-06, "log_odds_chosen": 1.4505621194839478, "log_odds_ratio": -0.3243867754936218, "logits/chosen": 393.34063720703125, "logits/rejected": 387.152099609375, "logps/chosen": -0.8490570783615112, "logps/rejected": -1.8110103607177734, "loss": 1.2897, "nll_loss": 1.1027252674102783, "rewards/accuracies": 0.8999999761581421, "rewards/chosen": -0.4245285391807556, "rewards/margins": 0.48097652196884155, "rewards/rejected": -0.9055051803588867, "step": 220 }, { "epoch": 1.3412816691505216, "grad_norm": 22.3747615814209, "learning_rate": 4.9999999999999996e-06, "log_odds_chosen": 1.0405768156051636, "log_odds_ratio": -0.37774857878685, "logits/chosen": 391.2264099121094, "logits/rejected": 382.2864685058594, "logps/chosen": -1.0447379350662231, "logps/rejected": -1.8071935176849365, "loss": 1.2781, "nll_loss": 1.2478437423706055, "rewards/accuracies": 0.875, "rewards/chosen": -0.5223689675331116, "rewards/margins": 0.38122788071632385, "rewards/rejected": -0.9035967588424683, "step": 225 }, { "epoch": 1.3710879284649775, "grad_norm": 22.041149139404297, "learning_rate": 4.945353550468404e-06, "log_odds_chosen": 2.0077407360076904, "log_odds_ratio": -0.25288665294647217, "logits/chosen": 370.07696533203125, "logits/rejected": 385.3541564941406, "logps/chosen": -0.90089350938797, "logps/rejected": -2.3661916255950928, "loss": 1.2539, "nll_loss": 0.989953339099884, "rewards/accuracies": 0.925000011920929, "rewards/chosen": -0.450446754693985, "rewards/margins": 0.7326492667198181, "rewards/rejected": -1.1830958127975464, "step": 230 }, { "epoch": 1.4008941877794336, "grad_norm": 33.446224212646484, "learning_rate": 4.8924605479008165e-06, "log_odds_chosen": 1.8689000606536865, "log_odds_ratio": -0.24617977440357208, "logits/chosen": 391.69732666015625, "logits/rejected": 402.88836669921875, "logps/chosen": -0.8436087369918823, "logps/rejected": -2.213329553604126, "loss": 1.2797, "nll_loss": 1.1102927923202515, "rewards/accuracies": 0.949999988079071, "rewards/chosen": -0.42180436849594116, "rewards/margins": 0.684860348701477, "rewards/rejected": -1.106664776802063, "step": 235 }, { "epoch": 1.4307004470938898, "grad_norm": 19.861835479736328, "learning_rate": 4.8412291827592715e-06, "log_odds_chosen": 1.848902940750122, "log_odds_ratio": -0.3047091066837311, "logits/chosen": 415.73486328125, "logits/rejected": 420.25909423828125, "logps/chosen": -0.8824371099472046, "logps/rejected": -2.3095035552978516, "loss": 1.2891, "nll_loss": 1.1334781646728516, "rewards/accuracies": 0.949999988079071, "rewards/chosen": -0.4412185549736023, "rewards/margins": 0.7135332226753235, "rewards/rejected": -1.1547517776489258, "step": 240 }, { "epoch": 1.4605067064083457, "grad_norm": 31.12167739868164, "learning_rate": 4.79157423749955e-06, "log_odds_chosen": 1.9949026107788086, "log_odds_ratio": -0.24574021995067596, "logits/chosen": 409.1098937988281, "logits/rejected": 394.39593505859375, "logps/chosen": -0.7639743089675903, "logps/rejected": -2.1869406700134277, "loss": 1.312, "nll_loss": 1.2122286558151245, "rewards/accuracies": 0.949999988079071, "rewards/chosen": -0.38198715448379517, "rewards/margins": 0.7114832997322083, "rewards/rejected": -1.0934703350067139, "step": 245 }, { "epoch": 1.4903129657228018, "grad_norm": 20.75017547607422, "learning_rate": 4.74341649025257e-06, "log_odds_chosen": 2.6609864234924316, "log_odds_ratio": -0.25564372539520264, "logits/chosen": 380.4820251464844, "logits/rejected": 385.7266845703125, "logps/chosen": -0.804089367389679, "logps/rejected": -2.9035263061523438, "loss": 1.2332, "nll_loss": 1.0138413906097412, "rewards/accuracies": 0.875, "rewards/chosen": -0.4020446836948395, "rewards/margins": 1.0497183799743652, "rewards/rejected": -1.4517631530761719, "step": 250 }, { "epoch": 1.520119225037258, "grad_norm": 26.518312454223633, "learning_rate": 4.696682183138622e-06, "log_odds_chosen": 2.0447866916656494, "log_odds_ratio": -0.22285214066505432, "logits/chosen": 399.64971923828125, "logits/rejected": 400.3153076171875, "logps/chosen": -0.8079016804695129, "logps/rejected": -2.2692599296569824, "loss": 1.2551, "nll_loss": 1.0453466176986694, "rewards/accuracies": 0.949999988079071, "rewards/chosen": -0.40395084023475647, "rewards/margins": 0.7306792140007019, "rewards/rejected": -1.1346299648284912, "step": 255 }, { "epoch": 1.5499254843517138, "grad_norm": 18.25040626525879, "learning_rate": 4.6513025470953165e-06, "log_odds_chosen": 1.7279441356658936, "log_odds_ratio": -0.3020802140235901, "logits/chosen": 380.86590576171875, "logits/rejected": 420.64007568359375, "logps/chosen": -0.8852326273918152, "logps/rejected": -2.209996461868286, "loss": 1.2132, "nll_loss": 1.1014916896820068, "rewards/accuracies": 0.8500000238418579, "rewards/chosen": -0.4426163136959076, "rewards/margins": 0.6623818278312683, "rewards/rejected": -1.104998230934143, "step": 260 }, { "epoch": 1.5797317436661698, "grad_norm": 17.494176864624023, "learning_rate": 4.607213376254634e-06, "log_odds_chosen": 1.6819546222686768, "log_odds_ratio": -0.3340403437614441, "logits/chosen": 388.51068115234375, "logits/rejected": 409.89971923828125, "logps/chosen": -0.8424839973449707, "logps/rejected": -2.041083574295044, "loss": 1.3572, "nll_loss": 1.1232020854949951, "rewards/accuracies": 0.875, "rewards/chosen": -0.42124199867248535, "rewards/margins": 0.5992996692657471, "rewards/rejected": -1.020541787147522, "step": 265 }, { "epoch": 1.6095380029806259, "grad_norm": 34.47447967529297, "learning_rate": 4.564354645876384e-06, "log_odds_chosen": 1.859960913658142, "log_odds_ratio": -0.2928032875061035, "logits/chosen": 372.35052490234375, "logits/rejected": 382.39117431640625, "logps/chosen": -0.9743008613586426, "logps/rejected": -2.432216167449951, "loss": 1.3166, "nll_loss": 1.270026683807373, "rewards/accuracies": 0.925000011920929, "rewards/chosen": -0.4871504306793213, "rewards/margins": 0.7289576530456543, "rewards/rejected": -1.2161080837249756, "step": 270 }, { "epoch": 1.639344262295082, "grad_norm": 21.639741897583008, "learning_rate": 4.522670168666455e-06, "log_odds_chosen": 1.463616967201233, "log_odds_ratio": -0.3027709126472473, "logits/chosen": 362.27044677734375, "logits/rejected": 350.851318359375, "logps/chosen": -0.7592640519142151, "logps/rejected": -1.6749483346939087, "loss": 1.2876, "nll_loss": 1.0598416328430176, "rewards/accuracies": 0.8500000238418579, "rewards/chosen": -0.37963202595710754, "rewards/margins": 0.45784205198287964, "rewards/rejected": -0.8374741673469543, "step": 275 }, { "epoch": 1.669150521609538, "grad_norm": 32.02434158325195, "learning_rate": 4.482107285003976e-06, "log_odds_chosen": 1.598225474357605, "log_odds_ratio": -0.3603498339653015, "logits/chosen": 410.98687744140625, "logits/rejected": 397.2725830078125, "logps/chosen": -0.8709872364997864, "logps/rejected": -1.9590165615081787, "loss": 1.285, "nll_loss": 1.1194555759429932, "rewards/accuracies": 0.824999988079071, "rewards/chosen": -0.4354936182498932, "rewards/margins": 0.5440146923065186, "rewards/rejected": -0.9795082807540894, "step": 280 }, { "epoch": 1.698956780923994, "grad_norm": 18.135896682739258, "learning_rate": 4.442616583193192e-06, "log_odds_chosen": 2.131542921066284, "log_odds_ratio": -0.2719998061656952, "logits/chosen": 375.759033203125, "logits/rejected": 377.7268981933594, "logps/chosen": -0.7118241786956787, "logps/rejected": -2.2590417861938477, "loss": 1.2113, "nll_loss": 1.054491400718689, "rewards/accuracies": 0.8999999761581421, "rewards/chosen": -0.35591208934783936, "rewards/margins": 0.7736088037490845, "rewards/rejected": -1.1295208930969238, "step": 285 }, { "epoch": 1.7287630402384502, "grad_norm": 16.36552619934082, "learning_rate": 4.404151646360276e-06, "log_odds_chosen": 2.539419651031494, "log_odds_ratio": -0.2885150611400604, "logits/chosen": 395.09832763671875, "logits/rejected": 429.5597229003906, "logps/chosen": -0.8963809013366699, "logps/rejected": -2.95475697517395, "loss": 1.2765, "nll_loss": 1.1113842725753784, "rewards/accuracies": 0.8999999761581421, "rewards/chosen": -0.44819045066833496, "rewards/margins": 1.0291881561279297, "rewards/rejected": -1.477378487586975, "step": 290 }, { "epoch": 1.758569299552906, "grad_norm": 20.132436752319336, "learning_rate": 4.366668823046865e-06, "log_odds_chosen": 1.553693413734436, "log_odds_ratio": -0.3133796155452728, "logits/chosen": 371.56475830078125, "logits/rejected": 413.505615234375, "logps/chosen": -0.9781205058097839, "logps/rejected": -2.0690226554870605, "loss": 1.2963, "nll_loss": 1.1234586238861084, "rewards/accuracies": 0.875, "rewards/chosen": -0.48906025290489197, "rewards/margins": 0.5454510450363159, "rewards/rejected": -1.0345113277435303, "step": 295 }, { "epoch": 1.788375558867362, "grad_norm": 26.103137969970703, "learning_rate": 4.330127018922194e-06, "log_odds_chosen": 2.1462044715881348, "log_odds_ratio": -0.24198386073112488, "logits/chosen": 409.97161865234375, "logits/rejected": 383.2239990234375, "logps/chosen": -0.7521153688430786, "logps/rejected": -2.3262906074523926, "loss": 1.1963, "nll_loss": 1.022304654121399, "rewards/accuracies": 0.925000011920929, "rewards/chosen": -0.3760576844215393, "rewards/margins": 0.7870877385139465, "rewards/rejected": -1.1631453037261963, "step": 300 }, { "epoch": 1.8181818181818183, "grad_norm": 20.291587829589844, "learning_rate": 4.294487507354011e-06, "log_odds_chosen": 1.8153495788574219, "log_odds_ratio": -0.3603833019733429, "logits/chosen": 399.9996032714844, "logits/rejected": 405.7534484863281, "logps/chosen": -1.029065728187561, "logps/rejected": -2.513669490814209, "loss": 1.2618, "nll_loss": 1.2645294666290283, "rewards/accuracies": 0.875, "rewards/chosen": -0.5145328640937805, "rewards/margins": 0.742301881313324, "rewards/rejected": -1.2568347454071045, "step": 305 }, { "epoch": 1.8479880774962743, "grad_norm": 19.601377487182617, "learning_rate": 4.259713756852986e-06, "log_odds_chosen": 3.238518238067627, "log_odds_ratio": -0.22569267451763153, "logits/chosen": 380.0532531738281, "logits/rejected": 386.49920654296875, "logps/chosen": -0.8288024067878723, "logps/rejected": -3.505189895629883, "loss": 1.3118, "nll_loss": 1.1829016208648682, "rewards/accuracies": 0.949999988079071, "rewards/chosen": -0.41440120339393616, "rewards/margins": 1.338193655014038, "rewards/rejected": -1.7525949478149414, "step": 310 }, { "epoch": 1.8777943368107302, "grad_norm": 19.882904052734375, "learning_rate": 4.225771273642583e-06, "log_odds_chosen": 1.2986266613006592, "log_odds_ratio": -0.3742119073867798, "logits/chosen": 354.4183044433594, "logits/rejected": 373.57757568359375, "logps/chosen": -1.0900706052780151, "logps/rejected": -2.065725326538086, "loss": 1.3407, "nll_loss": 1.3649649620056152, "rewards/accuracies": 0.8500000238418579, "rewards/chosen": -0.5450353026390076, "rewards/margins": 0.4878273606300354, "rewards/rejected": -1.032862663269043, "step": 315 }, { "epoch": 1.9076005961251863, "grad_norm": 21.79350471496582, "learning_rate": 4.192627457812106e-06, "log_odds_chosen": 2.3978288173675537, "log_odds_ratio": -0.24583116173744202, "logits/chosen": 379.4220886230469, "logits/rejected": 400.0596618652344, "logps/chosen": -0.7697328329086304, "logps/rejected": -2.6053688526153564, "loss": 1.1966, "nll_loss": 1.0758590698242188, "rewards/accuracies": 0.949999988079071, "rewards/chosen": -0.3848664164543152, "rewards/margins": 0.9178179502487183, "rewards/rejected": -1.3026844263076782, "step": 320 }, { "epoch": 1.9374068554396424, "grad_norm": 25.364540100097656, "learning_rate": 4.1602514716892185e-06, "log_odds_chosen": 2.511842727661133, "log_odds_ratio": -0.21811020374298096, "logits/chosen": 377.62139892578125, "logits/rejected": 409.9249572753906, "logps/chosen": -0.8184257745742798, "logps/rejected": -2.717911720275879, "loss": 1.2025, "nll_loss": 1.0009828805923462, "rewards/accuracies": 0.9750000238418579, "rewards/chosen": -0.4092128872871399, "rewards/margins": 0.9497429728507996, "rewards/rejected": -1.3589558601379395, "step": 325 }, { "epoch": 1.9672131147540983, "grad_norm": 24.96152114868164, "learning_rate": 4.128614119223853e-06, "log_odds_chosen": 1.425945520401001, "log_odds_ratio": -0.2635498642921448, "logits/chosen": 370.4853515625, "logits/rejected": 376.5760192871094, "logps/chosen": -0.8058730959892273, "logps/rejected": -1.770816445350647, "loss": 1.257, "nll_loss": 0.9558019638061523, "rewards/accuracies": 0.9750000238418579, "rewards/chosen": -0.40293654799461365, "rewards/margins": 0.48247161507606506, "rewards/rejected": -0.8854082226753235, "step": 330 }, { "epoch": 1.9970193740685542, "grad_norm": 17.690673828125, "learning_rate": 4.097687735311148e-06, "log_odds_chosen": 1.0468049049377441, "log_odds_ratio": -0.43120455741882324, "logits/chosen": 390.3360900878906, "logits/rejected": 377.1321105957031, "logps/chosen": -1.1694064140319824, "logps/rejected": -1.9213879108428955, "loss": 1.299, "nll_loss": 1.3618862628936768, "rewards/accuracies": 0.75, "rewards/chosen": -0.5847032070159912, "rewards/margins": 0.3759908080101013, "rewards/rejected": -0.9606939554214478, "step": 335 }, { "epoch": 1.9970193740685542, "eval_log_odds_chosen": 0.30739960074424744, "eval_log_odds_ratio": -0.6665536761283875, "eval_logits/chosen": 313.4974365234375, "eval_logits/rejected": 282.3182678222656, "eval_logps/chosen": -1.1013935804367065, "eval_logps/rejected": -1.3020689487457275, "eval_loss": 1.8186225891113281, "eval_nll_loss": 1.4682071208953857, "eval_rewards/accuracies": 0.5395683646202087, "eval_rewards/chosen": -0.5506967902183533, "eval_rewards/margins": 0.10033769905567169, "eval_rewards/rejected": -0.6510344743728638, "eval_runtime": 378.4505, "eval_samples_per_second": 1.461, "eval_steps_per_second": 0.367, "step": 335 }, { "epoch": 2.0268256333830106, "grad_norm": 19.424360275268555, "learning_rate": 4.0674460840998035e-06, "log_odds_chosen": 3.000514507293701, "log_odds_ratio": -0.10731367021799088, "logits/chosen": 394.8528747558594, "logits/rejected": 368.841796875, "logps/chosen": -0.4828576147556305, "logps/rejected": -2.412517547607422, "loss": 0.7587, "nll_loss": 0.8271492123603821, "rewards/accuracies": 0.9750000238418579, "rewards/chosen": -0.24142880737781525, "rewards/margins": 0.9648300409317017, "rewards/rejected": -1.206258773803711, "step": 340 }, { "epoch": 2.0566318926974665, "grad_norm": 13.576729774475098, "learning_rate": 4.037864265436242e-06, "log_odds_chosen": 3.940922498703003, "log_odds_ratio": -0.07164354622364044, "logits/chosen": 342.024658203125, "logits/rejected": 342.9552307128906, "logps/chosen": -0.48198527097702026, "logps/rejected": -3.1786067485809326, "loss": 0.6501, "nll_loss": 0.6376822590827942, "rewards/accuracies": 0.949999988079071, "rewards/chosen": -0.24099263548851013, "rewards/margins": 1.3483107089996338, "rewards/rejected": -1.5893033742904663, "step": 345 }, { "epoch": 2.0864381520119224, "grad_norm": 13.920299530029297, "learning_rate": 4.008918628686366e-06, "log_odds_chosen": 3.9218780994415283, "log_odds_ratio": -0.04757709056138992, "logits/chosen": 334.36187744140625, "logits/rejected": 374.3761291503906, "logps/chosen": -0.4404733180999756, "logps/rejected": -3.2984976768493652, "loss": 0.6334, "nll_loss": 0.6365537643432617, "rewards/accuracies": 1.0, "rewards/chosen": -0.2202366590499878, "rewards/margins": 1.4290122985839844, "rewards/rejected": -1.6492488384246826, "step": 350 }, { "epoch": 2.1162444113263787, "grad_norm": 12.827689170837402, "learning_rate": 3.980586693257065e-06, "log_odds_chosen": 4.072226047515869, "log_odds_ratio": -0.05848418548703194, "logits/chosen": 343.1537170410156, "logits/rejected": 358.5604248046875, "logps/chosen": -0.41472020745277405, "logps/rejected": -3.3263726234436035, "loss": 0.6337, "nll_loss": 0.6818060278892517, "rewards/accuracies": 1.0, "rewards/chosen": -0.20736010372638702, "rewards/margins": 1.4558260440826416, "rewards/rejected": -1.6631863117218018, "step": 355 }, { "epoch": 2.1460506706408347, "grad_norm": 14.782392501831055, "learning_rate": 3.952847075210474e-06, "log_odds_chosen": 4.195507049560547, "log_odds_ratio": -0.03957627713680267, "logits/chosen": 346.2444152832031, "logits/rejected": 393.8316955566406, "logps/chosen": -0.4012891352176666, "logps/rejected": -3.2268481254577637, "loss": 0.62, "nll_loss": 0.5778239965438843, "rewards/accuracies": 1.0, "rewards/chosen": -0.2006445676088333, "rewards/margins": 1.4127795696258545, "rewards/rejected": -1.6134240627288818, "step": 360 }, { "epoch": 2.1758569299552906, "grad_norm": 13.0780029296875, "learning_rate": 3.925679419426603e-06, "log_odds_chosen": 3.8077125549316406, "log_odds_ratio": -0.06930537521839142, "logits/chosen": 345.62933349609375, "logits/rejected": 324.48724365234375, "logps/chosen": -0.416290283203125, "logps/rejected": -3.0446524620056152, "loss": 0.6211, "nll_loss": 0.5697866678237915, "rewards/accuracies": 0.9750000238418579, "rewards/chosen": -0.2081451416015625, "rewards/margins": 1.3141810894012451, "rewards/rejected": -1.5223262310028076, "step": 365 }, { "epoch": 2.2056631892697465, "grad_norm": 13.796915054321289, "learning_rate": 3.8990643368252725e-06, "log_odds_chosen": 3.758181095123291, "log_odds_ratio": -0.08863835036754608, "logits/chosen": 309.7760009765625, "logits/rejected": 327.28875732421875, "logps/chosen": -0.4810813367366791, "logps/rejected": -3.211531400680542, "loss": 0.6805, "nll_loss": 0.6794728636741638, "rewards/accuracies": 0.9750000238418579, "rewards/chosen": -0.24054066836833954, "rewards/margins": 1.365225076675415, "rewards/rejected": -1.605765700340271, "step": 370 }, { "epoch": 2.235469448584203, "grad_norm": 12.518247604370117, "learning_rate": 3.872983346207417e-06, "log_odds_chosen": 3.9930922985076904, "log_odds_ratio": -0.07049717009067535, "logits/chosen": 364.5960388183594, "logits/rejected": 381.1838684082031, "logps/chosen": -0.47075748443603516, "logps/rejected": -3.3296546936035156, "loss": 0.6191, "nll_loss": 0.6227518320083618, "rewards/accuracies": 0.9750000238418579, "rewards/chosen": -0.23537874221801758, "rewards/margins": 1.4294484853744507, "rewards/rejected": -1.6648273468017578, "step": 375 }, { "epoch": 2.2652757078986587, "grad_norm": 23.583436965942383, "learning_rate": 3.847418820319329e-06, "log_odds_chosen": 3.55949330329895, "log_odds_ratio": -0.048723362386226654, "logits/chosen": 364.0585632324219, "logits/rejected": 337.66424560546875, "logps/chosen": -0.41810473799705505, "logps/rejected": -2.7338831424713135, "loss": 0.6367, "nll_loss": 0.6199051737785339, "rewards/accuracies": 1.0, "rewards/chosen": -0.20905236899852753, "rewards/margins": 1.1578892469406128, "rewards/rejected": -1.3669415712356567, "step": 380 }, { "epoch": 2.2950819672131146, "grad_norm": 12.274104118347168, "learning_rate": 3.822353935782191e-06, "log_odds_chosen": 4.226048946380615, "log_odds_ratio": -0.04617181420326233, "logits/chosen": 359.08154296875, "logits/rejected": 344.7435607910156, "logps/chosen": -0.36728304624557495, "logps/rejected": -3.1797947883605957, "loss": 0.5982, "nll_loss": 0.536335289478302, "rewards/accuracies": 0.9750000238418579, "rewards/chosen": -0.18364152312278748, "rewards/margins": 1.4062559604644775, "rewards/rejected": -1.5898973941802979, "step": 385 }, { "epoch": 2.3248882265275705, "grad_norm": 15.543499946594238, "learning_rate": 3.79777262656375e-06, "log_odds_chosen": 3.8094513416290283, "log_odds_ratio": -0.03931097313761711, "logits/chosen": 373.632568359375, "logits/rejected": 393.21295166015625, "logps/chosen": -0.46378618478775024, "logps/rejected": -3.2327797412872314, "loss": 0.667, "nll_loss": 0.6460134387016296, "rewards/accuracies": 1.0, "rewards/chosen": -0.23189309239387512, "rewards/margins": 1.3844965696334839, "rewards/rejected": -1.6163898706436157, "step": 390 }, { "epoch": 2.354694485842027, "grad_norm": 13.162529945373535, "learning_rate": 3.773659540699707e-06, "log_odds_chosen": 3.9874942302703857, "log_odds_ratio": -0.05443572998046875, "logits/chosen": 368.2042236328125, "logits/rejected": 346.73590087890625, "logps/chosen": -0.3340614438056946, "logps/rejected": -2.9790287017822266, "loss": 0.5966, "nll_loss": 0.5552054643630981, "rewards/accuracies": 1.0, "rewards/chosen": -0.1670307219028473, "rewards/margins": 1.3224836587905884, "rewards/rejected": -1.4895143508911133, "step": 395 }, { "epoch": 2.384500745156483, "grad_norm": 11.184009552001953, "learning_rate": 3.75e-06, "log_odds_chosen": 4.703558444976807, "log_odds_ratio": -0.06566660106182098, "logits/chosen": 368.7803955078125, "logits/rejected": 383.49188232421875, "logps/chosen": -0.4126485288143158, "logps/rejected": -3.8882851600646973, "loss": 0.619, "nll_loss": 0.5621041059494019, "rewards/accuracies": 0.949999988079071, "rewards/chosen": -0.2063242644071579, "rewards/margins": 1.7378184795379639, "rewards/rejected": -1.9441425800323486, "step": 400 }, { "epoch": 2.4143070044709387, "grad_norm": 12.980825424194336, "learning_rate": 3.72677996249965e-06, "log_odds_chosen": 4.317046165466309, "log_odds_ratio": -0.07360969483852386, "logits/chosen": 422.6468200683594, "logits/rejected": 380.22998046875, "logps/chosen": -0.43779826164245605, "logps/rejected": -3.444517135620117, "loss": 0.6456, "nll_loss": 0.6048913598060608, "rewards/accuracies": 0.9750000238418579, "rewards/chosen": -0.21889913082122803, "rewards/margins": 1.5033594369888306, "rewards/rejected": -1.7222585678100586, "step": 405 }, { "epoch": 2.444113263785395, "grad_norm": 11.386286735534668, "learning_rate": 3.7039859874359616e-06, "log_odds_chosen": 3.6852588653564453, "log_odds_ratio": -0.06955203413963318, "logits/chosen": 365.75726318359375, "logits/rejected": 373.08935546875, "logps/chosen": -0.4394207000732422, "logps/rejected": -2.8892974853515625, "loss": 0.6307, "nll_loss": 0.6479864120483398, "rewards/accuracies": 0.9750000238418579, "rewards/chosen": -0.2197103500366211, "rewards/margins": 1.2249382734298706, "rewards/rejected": -1.4446487426757812, "step": 410 }, { "epoch": 2.473919523099851, "grad_norm": 18.22783088684082, "learning_rate": 3.68160520255362e-06, "log_odds_chosen": 4.396167755126953, "log_odds_ratio": -0.04461110755801201, "logits/chosen": 371.11077880859375, "logits/rejected": 393.236328125, "logps/chosen": -0.4007198214530945, "logps/rejected": -3.539832353591919, "loss": 0.6134, "nll_loss": 0.5493707656860352, "rewards/accuracies": 1.0, "rewards/chosen": -0.20035991072654724, "rewards/margins": 1.5695561170578003, "rewards/rejected": -1.7699161767959595, "step": 415 }, { "epoch": 2.503725782414307, "grad_norm": 12.961064338684082, "learning_rate": 3.6596252735569993e-06, "log_odds_chosen": 3.7051539421081543, "log_odds_ratio": -0.06346116960048676, "logits/chosen": 368.0414123535156, "logits/rejected": 338.7422180175781, "logps/chosen": -0.38283970952033997, "logps/rejected": -2.9152991771698, "loss": 0.6314, "nll_loss": 0.6913596987724304, "rewards/accuracies": 0.9750000238418579, "rewards/chosen": -0.19141985476016998, "rewards/margins": 1.2662298679351807, "rewards/rejected": -1.4576495885849, "step": 420 }, { "epoch": 2.533532041728763, "grad_norm": 14.195425033569336, "learning_rate": 3.6380343755449946e-06, "log_odds_chosen": 4.560431480407715, "log_odds_ratio": -0.03129279986023903, "logits/chosen": 362.96746826171875, "logits/rejected": 349.57958984375, "logps/chosen": -0.4239065647125244, "logps/rejected": -3.7524616718292236, "loss": 0.6005, "nll_loss": 0.6241891980171204, "rewards/accuracies": 1.0, "rewards/chosen": -0.2119532823562622, "rewards/margins": 1.66427743434906, "rewards/rejected": -1.8762308359146118, "step": 425 }, { "epoch": 2.563338301043219, "grad_norm": 13.435834884643555, "learning_rate": 3.6168211662780912e-06, "log_odds_chosen": 4.4234299659729, "log_odds_ratio": -0.03766634315252304, "logits/chosen": 361.0891418457031, "logits/rejected": 362.5020446777344, "logps/chosen": -0.3212960362434387, "logps/rejected": -3.3532252311706543, "loss": 0.6194, "nll_loss": 0.5339005589485168, "rewards/accuracies": 1.0, "rewards/chosen": -0.16064801812171936, "rewards/margins": 1.5159647464752197, "rewards/rejected": -1.6766126155853271, "step": 430 }, { "epoch": 2.593144560357675, "grad_norm": 13.10000228881836, "learning_rate": 3.5959747611403815e-06, "log_odds_chosen": 3.667933702468872, "log_odds_ratio": -0.049441076815128326, "logits/chosen": 370.7209777832031, "logits/rejected": 329.81048583984375, "logps/chosen": -0.3787091374397278, "logps/rejected": -2.830317735671997, "loss": 0.6284, "nll_loss": 0.5626553297042847, "rewards/accuracies": 1.0, "rewards/chosen": -0.1893545687198639, "rewards/margins": 1.2258042097091675, "rewards/rejected": -1.4151588678359985, "step": 435 }, { "epoch": 2.6229508196721314, "grad_norm": 16.336698532104492, "learning_rate": 3.575484709670971e-06, "log_odds_chosen": 4.056857585906982, "log_odds_ratio": -0.035340312868356705, "logits/chosen": 350.6004333496094, "logits/rejected": 356.38323974609375, "logps/chosen": -0.3078530430793762, "logps/rejected": -2.8841094970703125, "loss": 0.6605, "nll_loss": 0.6005590558052063, "rewards/accuracies": 1.0, "rewards/chosen": -0.1539265215396881, "rewards/margins": 1.288128137588501, "rewards/rejected": -1.4420547485351562, "step": 440 }, { "epoch": 2.6527570789865873, "grad_norm": 16.911279678344727, "learning_rate": 3.5553409735498294e-06, "log_odds_chosen": 4.540930271148682, "log_odds_ratio": -0.021845344454050064, "logits/chosen": 333.2252197265625, "logits/rejected": 360.4376525878906, "logps/chosen": -0.3848707973957062, "logps/rejected": -3.6115005016326904, "loss": 0.6305, "nll_loss": 0.616925060749054, "rewards/accuracies": 1.0, "rewards/chosen": -0.1924353986978531, "rewards/margins": 1.6133148670196533, "rewards/rejected": -1.8057502508163452, "step": 445 }, { "epoch": 2.682563338301043, "grad_norm": 12.66332721710205, "learning_rate": 3.535533905932738e-06, "log_odds_chosen": 3.8172874450683594, "log_odds_ratio": -0.04109153896570206, "logits/chosen": 363.5258483886719, "logits/rejected": 355.9951171875, "logps/chosen": -0.442394495010376, "logps/rejected": -3.0551743507385254, "loss": 0.6594, "nll_loss": 0.6155456304550171, "rewards/accuracies": 1.0, "rewards/chosen": -0.221197247505188, "rewards/margins": 1.3063898086547852, "rewards/rejected": -1.5275871753692627, "step": 450 }, { "epoch": 2.712369597615499, "grad_norm": 11.626214981079102, "learning_rate": 3.516054232038716e-06, "log_odds_chosen": 3.7958292961120605, "log_odds_ratio": -0.05676658824086189, "logits/chosen": 394.8970947265625, "logits/rejected": 370.40692138671875, "logps/chosen": -0.49928539991378784, "logps/rejected": -3.0073413848876953, "loss": 0.6971, "nll_loss": 0.6259102821350098, "rewards/accuracies": 0.9750000238418579, "rewards/chosen": -0.24964269995689392, "rewards/margins": 1.2540279626846313, "rewards/rejected": -1.5036706924438477, "step": 455 }, { "epoch": 2.742175856929955, "grad_norm": 12.148904800415039, "learning_rate": 3.496893030901177e-06, "log_odds_chosen": 4.385091781616211, "log_odds_ratio": -0.0362153984606266, "logits/chosen": 390.84942626953125, "logits/rejected": 403.558837890625, "logps/chosen": -0.35050949454307556, "logps/rejected": -3.5067138671875, "loss": 0.64, "nll_loss": 0.6797647476196289, "rewards/accuracies": 1.0, "rewards/chosen": -0.17525474727153778, "rewards/margins": 1.5781023502349854, "rewards/rejected": -1.75335693359375, "step": 460 }, { "epoch": 2.7719821162444114, "grad_norm": 16.0198974609375, "learning_rate": 3.478041718201262e-06, "log_odds_chosen": 4.281471252441406, "log_odds_ratio": -0.04437633976340294, "logits/chosen": 379.65142822265625, "logits/rejected": 373.5622863769531, "logps/chosen": -0.3376389443874359, "logps/rejected": -3.2862961292266846, "loss": 0.6475, "nll_loss": 0.6045721769332886, "rewards/accuracies": 1.0, "rewards/chosen": -0.16881947219371796, "rewards/margins": 1.4743287563323975, "rewards/rejected": -1.6431480646133423, "step": 465 }, { "epoch": 2.8017883755588673, "grad_norm": 20.6240291595459, "learning_rate": 3.459492030108319e-06, "log_odds_chosen": 3.6953399181365967, "log_odds_ratio": -0.07698026299476624, "logits/chosen": 384.45758056640625, "logits/rejected": 344.5814208984375, "logps/chosen": -0.43422994017601013, "logps/rejected": -2.9317374229431152, "loss": 0.6539, "nll_loss": 0.6051168441772461, "rewards/accuracies": 0.9750000238418579, "rewards/chosen": -0.21711497008800507, "rewards/margins": 1.2487537860870361, "rewards/rejected": -1.4658687114715576, "step": 470 }, { "epoch": 2.8315946348733236, "grad_norm": 16.241731643676758, "learning_rate": 3.441236008058426e-06, "log_odds_chosen": 4.172859191894531, "log_odds_ratio": -0.03734353557229042, "logits/chosen": 359.6179504394531, "logits/rejected": 372.67559814453125, "logps/chosen": -0.3877125382423401, "logps/rejected": -3.2986769676208496, "loss": 0.6017, "nll_loss": 0.5468435287475586, "rewards/accuracies": 1.0, "rewards/chosen": -0.19385626912117004, "rewards/margins": 1.4554822444915771, "rewards/rejected": -1.6493384838104248, "step": 475 }, { "epoch": 2.8614008941877795, "grad_norm": 13.836338996887207, "learning_rate": 3.4232659844072883e-06, "log_odds_chosen": 5.078559875488281, "log_odds_ratio": -0.040774669498205185, "logits/chosen": 369.23388671875, "logits/rejected": 398.89654541015625, "logps/chosen": -0.4116280674934387, "logps/rejected": -4.2671427726745605, "loss": 0.5823, "nll_loss": 0.5545191764831543, "rewards/accuracies": 1.0, "rewards/chosen": -0.20581403374671936, "rewards/margins": 1.9277572631835938, "rewards/rejected": -2.1335713863372803, "step": 480 }, { "epoch": 2.8912071535022354, "grad_norm": 13.629563331604004, "learning_rate": 3.4055745688987482e-06, "log_odds_chosen": 3.756863832473755, "log_odds_ratio": -0.05358726903796196, "logits/chosen": 360.4635925292969, "logits/rejected": 379.5230407714844, "logps/chosen": -0.4683550000190735, "logps/rejected": -3.0478286743164062, "loss": 0.6356, "nll_loss": 0.5991590023040771, "rewards/accuracies": 0.9750000238418579, "rewards/chosen": -0.23417750000953674, "rewards/margins": 1.2897369861602783, "rewards/rejected": -1.5239143371582031, "step": 485 }, { "epoch": 2.9210134128166914, "grad_norm": 11.519225120544434, "learning_rate": 3.3881546358946924e-06, "log_odds_chosen": 4.434014320373535, "log_odds_ratio": -0.048062875866889954, "logits/chosen": 385.712158203125, "logits/rejected": 336.68524169921875, "logps/chosen": -0.45937252044677734, "logps/rejected": -3.5976157188415527, "loss": 0.6968, "nll_loss": 0.5958720445632935, "rewards/accuracies": 1.0, "rewards/chosen": -0.22968626022338867, "rewards/margins": 1.5691215991973877, "rewards/rejected": -1.7988078594207764, "step": 490 }, { "epoch": 2.9508196721311473, "grad_norm": 16.139705657958984, "learning_rate": 3.37099931231621e-06, "log_odds_chosen": 4.680273532867432, "log_odds_ratio": -0.062451474368572235, "logits/chosen": 375.66119384765625, "logits/rejected": 394.88690185546875, "logps/chosen": -0.4953223764896393, "logps/rejected": -3.950305938720703, "loss": 0.6257, "nll_loss": 0.628521203994751, "rewards/accuracies": 0.9750000238418579, "rewards/chosen": -0.24766118824481964, "rewards/margins": 1.7274919748306274, "rewards/rejected": -1.9751529693603516, "step": 495 }, { "epoch": 2.9806259314456036, "grad_norm": 15.226314544677734, "learning_rate": 3.3541019662496845e-06, "log_odds_chosen": 4.290003776550293, "log_odds_ratio": -0.04303581267595291, "logits/chosen": 379.34765625, "logits/rejected": 365.36773681640625, "logps/chosen": -0.3512369692325592, "logps/rejected": -3.2438788414001465, "loss": 0.6379, "nll_loss": 0.5231742858886719, "rewards/accuracies": 1.0, "rewards/chosen": -0.1756184846162796, "rewards/margins": 1.44632089138031, "rewards/rejected": -1.6219394207000732, "step": 500 }, { "epoch": 2.9865871833084947, "eval_log_odds_chosen": 0.33574143052101135, "eval_log_odds_ratio": -0.6769095063209534, "eval_logits/chosen": 292.74676513671875, "eval_logits/rejected": 259.2041320800781, "eval_logps/chosen": -1.2910676002502441, "eval_logps/rejected": -1.5239651203155518, "eval_loss": 2.003471851348877, "eval_nll_loss": 1.6439613103866577, "eval_rewards/accuracies": 0.6115108132362366, "eval_rewards/chosen": -0.6455338001251221, "eval_rewards/margins": 0.1164487823843956, "eval_rewards/rejected": -0.7619825601577759, "eval_runtime": 378.0521, "eval_samples_per_second": 1.463, "eval_steps_per_second": 0.368, "step": 501 }, { "epoch": 2.9865871833084947, "step": 501, "total_flos": 0.0, "train_loss": 1.6206341998543805, "train_runtime": 33576.7022, "train_samples_per_second": 0.479, "train_steps_per_second": 0.015 } ], "logging_steps": 5, "max_steps": 501, "num_input_tokens_seen": 0, "num_train_epochs": 3, "save_steps": 500, "stateful_callbacks": { "TrainerControl": { "args": { "should_epoch_stop": false, "should_evaluate": false, "should_log": false, "should_save": false, "should_training_stop": false }, "attributes": {} } }, "total_flos": 0.0, "train_batch_size": 2, "trial_name": null, "trial_params": null }