zephyr-7b-sft-full-orpo / trainer_state.json
statking's picture
Model save
473666a verified
raw
history blame
63.9 kB
{
"best_metric": null,
"best_model_checkpoint": null,
"epoch": 0.9994756161510225,
"eval_steps": 500,
"global_step": 953,
"is_hyper_param_search": false,
"is_local_process_zero": true,
"is_world_process_zero": true,
"log_history": [
{
"epoch": 0.01048767697954903,
"grad_norm": 12.504458138350461,
"learning_rate": 2.0000000000000003e-06,
"log_odds_chosen": 0.1660214066505432,
"log_odds_ratio": -0.6960338354110718,
"logits/chosen": -2.542905330657959,
"logits/rejected": -2.5316882133483887,
"logps/chosen": -0.9998037219047546,
"logps/rejected": -1.0999689102172852,
"loss": 2.7433,
"nll_loss": 2.6550583839416504,
"rewards/accuracies": 0.550000011920929,
"rewards/chosen": -0.04999018833041191,
"rewards/margins": 0.005008256994187832,
"rewards/rejected": -0.05499844625592232,
"step": 10
},
{
"epoch": 0.02097535395909806,
"grad_norm": 3.296398746092505,
"learning_rate": 4.000000000000001e-06,
"log_odds_chosen": 0.1942831575870514,
"log_odds_ratio": -0.6660380959510803,
"logits/chosen": -3.148456335067749,
"logits/rejected": -3.171660900115967,
"logps/chosen": -0.7626909613609314,
"logps/rejected": -0.8731427192687988,
"loss": 0.563,
"nll_loss": 0.5225270986557007,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.03813454881310463,
"rewards/margins": 0.00552258500829339,
"rewards/rejected": -0.04365713149309158,
"step": 20
},
{
"epoch": 0.03146303093864709,
"grad_norm": 2.4400188978085695,
"learning_rate": 6e-06,
"log_odds_chosen": 0.2339784801006317,
"log_odds_ratio": -0.6537522673606873,
"logits/chosen": -2.9630327224731445,
"logits/rejected": -2.9368481636047363,
"logps/chosen": -0.8345462679862976,
"logps/rejected": -0.9655241966247559,
"loss": 0.5355,
"nll_loss": 0.4940575659275055,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.04172731190919876,
"rewards/margins": 0.0065488978289067745,
"rewards/rejected": -0.04827621206641197,
"step": 30
},
{
"epoch": 0.04195070791819612,
"grad_norm": 2.765802378357493,
"learning_rate": 8.000000000000001e-06,
"log_odds_chosen": 0.15870003402233124,
"log_odds_ratio": -0.6969180107116699,
"logits/chosen": -2.8065195083618164,
"logits/rejected": -2.7910008430480957,
"logps/chosen": -0.8027766346931458,
"logps/rejected": -0.9165509343147278,
"loss": 0.5199,
"nll_loss": 0.48035889863967896,
"rewards/accuracies": 0.5625,
"rewards/chosen": -0.04013883322477341,
"rewards/margins": 0.005688714794814587,
"rewards/rejected": -0.04582754150032997,
"step": 40
},
{
"epoch": 0.05243838489774515,
"grad_norm": 2.7404814506796704,
"learning_rate": 1e-05,
"log_odds_chosen": 0.24872338771820068,
"log_odds_ratio": -0.680080771446228,
"logits/chosen": -2.7704856395721436,
"logits/rejected": -2.77298641204834,
"logps/chosen": -0.7987793684005737,
"logps/rejected": -0.9668463468551636,
"loss": 0.5424,
"nll_loss": 0.48421746492385864,
"rewards/accuracies": 0.59375,
"rewards/chosen": -0.03993896767497063,
"rewards/margins": 0.00840335339307785,
"rewards/rejected": -0.048342324793338776,
"step": 50
},
{
"epoch": 0.06292606187729417,
"grad_norm": 2.7601739927853473,
"learning_rate": 1.2e-05,
"log_odds_chosen": 0.21160352230072021,
"log_odds_ratio": -0.6764382123947144,
"logits/chosen": -3.0032615661621094,
"logits/rejected": -2.9960169792175293,
"logps/chosen": -0.7965995669364929,
"logps/rejected": -0.917363166809082,
"loss": 0.5463,
"nll_loss": 0.516124427318573,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.039829984307289124,
"rewards/margins": 0.006038171239197254,
"rewards/rejected": -0.045868150889873505,
"step": 60
},
{
"epoch": 0.07341373885684321,
"grad_norm": 3.2123267767300128,
"learning_rate": 1.4e-05,
"log_odds_chosen": 0.19886036217212677,
"log_odds_ratio": -0.690485417842865,
"logits/chosen": -2.978163719177246,
"logits/rejected": -3.0078656673431396,
"logps/chosen": -0.8206535577774048,
"logps/rejected": -0.9310994148254395,
"loss": 0.5403,
"nll_loss": 0.530234694480896,
"rewards/accuracies": 0.574999988079071,
"rewards/chosen": -0.04103267565369606,
"rewards/margins": 0.0055222949013113976,
"rewards/rejected": -0.046554967761039734,
"step": 70
},
{
"epoch": 0.08390141583639224,
"grad_norm": 3.267750524500123,
"learning_rate": 1.6000000000000003e-05,
"log_odds_chosen": 0.1725669652223587,
"log_odds_ratio": -0.689757764339447,
"logits/chosen": -2.963442087173462,
"logits/rejected": -2.953914165496826,
"logps/chosen": -0.8903671503067017,
"logps/rejected": -1.0184500217437744,
"loss": 0.5632,
"nll_loss": 0.48384732007980347,
"rewards/accuracies": 0.518750011920929,
"rewards/chosen": -0.0445183590054512,
"rewards/margins": 0.006404136773198843,
"rewards/rejected": -0.050922494381666183,
"step": 80
},
{
"epoch": 0.09438909281594127,
"grad_norm": 6.338896835312273,
"learning_rate": 1.8e-05,
"log_odds_chosen": 0.2590278387069702,
"log_odds_ratio": -0.6696828603744507,
"logits/chosen": -2.7556283473968506,
"logits/rejected": -2.759223461151123,
"logps/chosen": -0.8806008100509644,
"logps/rejected": -1.0427037477493286,
"loss": 0.5599,
"nll_loss": 0.49117976427078247,
"rewards/accuracies": 0.59375,
"rewards/chosen": -0.044030044227838516,
"rewards/margins": 0.008105142042040825,
"rewards/rejected": -0.05213518068194389,
"step": 90
},
{
"epoch": 0.1048767697954903,
"grad_norm": 2.844482964932932,
"learning_rate": 2e-05,
"log_odds_chosen": 0.20001336932182312,
"log_odds_ratio": -0.6672823429107666,
"logits/chosen": -2.836613178253174,
"logits/rejected": -2.826347827911377,
"logps/chosen": -0.8816211819648743,
"logps/rejected": -1.0050264596939087,
"loss": 0.5675,
"nll_loss": 0.5239149332046509,
"rewards/accuracies": 0.574999988079071,
"rewards/chosen": -0.044081058353185654,
"rewards/margins": 0.006170268170535564,
"rewards/rejected": -0.05025132745504379,
"step": 100
},
{
"epoch": 0.11536444677503933,
"grad_norm": 2.717573270122186,
"learning_rate": 1.9069251784911845e-05,
"log_odds_chosen": 0.26770642399787903,
"log_odds_ratio": -0.6399692296981812,
"logits/chosen": -2.8041529655456543,
"logits/rejected": -2.828374147415161,
"logps/chosen": -0.8482567071914673,
"logps/rejected": -1.021328330039978,
"loss": 0.568,
"nll_loss": 0.5094035863876343,
"rewards/accuracies": 0.625,
"rewards/chosen": -0.042412832379341125,
"rewards/margins": 0.008653589524328709,
"rewards/rejected": -0.05106641724705696,
"step": 110
},
{
"epoch": 0.12585212375458835,
"grad_norm": 2.3522582585650906,
"learning_rate": 1.825741858350554e-05,
"log_odds_chosen": 0.2770318388938904,
"log_odds_ratio": -0.6538770198822021,
"logits/chosen": -2.9046432971954346,
"logits/rejected": -2.921250343322754,
"logps/chosen": -0.8698671460151672,
"logps/rejected": -1.0593181848526,
"loss": 0.6048,
"nll_loss": 0.5620476007461548,
"rewards/accuracies": 0.6312500238418579,
"rewards/chosen": -0.0434933602809906,
"rewards/margins": 0.009472550824284554,
"rewards/rejected": -0.05296590179204941,
"step": 120
},
{
"epoch": 0.1363398007341374,
"grad_norm": 2.3512564845307704,
"learning_rate": 1.7541160386140587e-05,
"log_odds_chosen": 0.213302880525589,
"log_odds_ratio": -0.6861675977706909,
"logits/chosen": -2.926781177520752,
"logits/rejected": -2.930361747741699,
"logps/chosen": -0.9192083477973938,
"logps/rejected": -1.06519615650177,
"loss": 0.5923,
"nll_loss": 0.5574383735656738,
"rewards/accuracies": 0.5625,
"rewards/chosen": -0.04596042260527611,
"rewards/margins": 0.007299385964870453,
"rewards/rejected": -0.05325980857014656,
"step": 130
},
{
"epoch": 0.14682747771368643,
"grad_norm": 2.2489368047485705,
"learning_rate": 1.6903085094570334e-05,
"log_odds_chosen": 0.24789170920848846,
"log_odds_ratio": -0.655090868473053,
"logits/chosen": -2.9084389209747314,
"logits/rejected": -2.9173099994659424,
"logps/chosen": -0.9441210031509399,
"logps/rejected": -1.1045926809310913,
"loss": 0.5882,
"nll_loss": 0.5544429421424866,
"rewards/accuracies": 0.574999988079071,
"rewards/chosen": -0.047206051647663116,
"rewards/margins": 0.008023588918149471,
"rewards/rejected": -0.05522964149713516,
"step": 140
},
{
"epoch": 0.15731515469323545,
"grad_norm": 2.6715309670512903,
"learning_rate": 1.6329931618554523e-05,
"log_odds_chosen": 0.14654028415679932,
"log_odds_ratio": -0.7416929006576538,
"logits/chosen": -2.8286139965057373,
"logits/rejected": -2.842860698699951,
"logps/chosen": -0.9699670672416687,
"logps/rejected": -1.0669214725494385,
"loss": 0.5441,
"nll_loss": 0.5359360575675964,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.048498354852199554,
"rewards/margins": 0.004847715608775616,
"rewards/rejected": -0.053346067667007446,
"step": 150
},
{
"epoch": 0.16780283167278448,
"grad_norm": 2.4917874181934616,
"learning_rate": 1.5811388300841898e-05,
"log_odds_chosen": 0.19475655257701874,
"log_odds_ratio": -0.664051353931427,
"logits/chosen": -2.8252522945404053,
"logits/rejected": -2.839994192123413,
"logps/chosen": -0.9179447889328003,
"logps/rejected": -1.0352815389633179,
"loss": 0.6078,
"nll_loss": 0.5540346503257751,
"rewards/accuracies": 0.6312500238418579,
"rewards/chosen": -0.045897237956523895,
"rewards/margins": 0.005866840481758118,
"rewards/rejected": -0.05176408216357231,
"step": 160
},
{
"epoch": 0.1782905086523335,
"grad_norm": 2.493896039254152,
"learning_rate": 1.533929977694741e-05,
"log_odds_chosen": 0.25445470213890076,
"log_odds_ratio": -0.6574397087097168,
"logits/chosen": -2.895998477935791,
"logits/rejected": -2.9125123023986816,
"logps/chosen": -0.8917832374572754,
"logps/rejected": -1.0586717128753662,
"loss": 0.5884,
"nll_loss": 0.5544494986534119,
"rewards/accuracies": 0.581250011920929,
"rewards/chosen": -0.04458915814757347,
"rewards/margins": 0.008344428613781929,
"rewards/rejected": -0.05293358489871025,
"step": 170
},
{
"epoch": 0.18877818563188253,
"grad_norm": 2.368451448201635,
"learning_rate": 1.49071198499986e-05,
"log_odds_chosen": 0.2552924156188965,
"log_odds_ratio": -0.6543556451797485,
"logits/chosen": -2.8886399269104004,
"logits/rejected": -2.905686378479004,
"logps/chosen": -0.9206914901733398,
"logps/rejected": -1.091048240661621,
"loss": 0.5686,
"nll_loss": 0.551173985004425,
"rewards/accuracies": 0.606249988079071,
"rewards/chosen": -0.04603457450866699,
"rewards/margins": 0.008517834357917309,
"rewards/rejected": -0.054552413523197174,
"step": 180
},
{
"epoch": 0.19926586261143156,
"grad_norm": 4.734046585912702,
"learning_rate": 1.4509525002200235e-05,
"log_odds_chosen": 0.21173310279846191,
"log_odds_ratio": -0.6579927206039429,
"logits/chosen": -2.9355111122131348,
"logits/rejected": -2.952430009841919,
"logps/chosen": -0.9388859868049622,
"logps/rejected": -1.0733187198638916,
"loss": 0.5936,
"nll_loss": 0.6142745018005371,
"rewards/accuracies": 0.606249988079071,
"rewards/chosen": -0.04694430157542229,
"rewards/margins": 0.006721635349094868,
"rewards/rejected": -0.05366594344377518,
"step": 190
},
{
"epoch": 0.2097535395909806,
"grad_norm": 2.2391424397427073,
"learning_rate": 1.4142135623730951e-05,
"log_odds_chosen": 0.28418153524398804,
"log_odds_ratio": -0.6668760180473328,
"logits/chosen": -2.873599052429199,
"logits/rejected": -2.9066414833068848,
"logps/chosen": -0.9204713702201843,
"logps/rejected": -1.128112554550171,
"loss": 0.5689,
"nll_loss": 0.5723541975021362,
"rewards/accuracies": 0.581250011920929,
"rewards/chosen": -0.046023570001125336,
"rewards/margins": 0.010382059030234814,
"rewards/rejected": -0.056405626237392426,
"step": 200
},
{
"epoch": 0.22024121657052964,
"grad_norm": 2.1684330770876152,
"learning_rate": 1.3801311186847084e-05,
"log_odds_chosen": 0.11919783055782318,
"log_odds_ratio": -0.7173447012901306,
"logits/chosen": -2.884079933166504,
"logits/rejected": -2.8981668949127197,
"logps/chosen": -0.8726099729537964,
"logps/rejected": -0.9488958120346069,
"loss": 0.5693,
"nll_loss": 0.5325449109077454,
"rewards/accuracies": 0.606249988079071,
"rewards/chosen": -0.04363049939274788,
"rewards/margins": 0.0038142912089824677,
"rewards/rejected": -0.04744479060173035,
"step": 210
},
{
"epoch": 0.23072889355007867,
"grad_norm": 2.510753834710904,
"learning_rate": 1.3483997249264842e-05,
"log_odds_chosen": 0.18100012838840485,
"log_odds_ratio": -0.7047401666641235,
"logits/chosen": -2.8885810375213623,
"logits/rejected": -2.8980116844177246,
"logps/chosen": -0.8880792856216431,
"logps/rejected": -1.0071966648101807,
"loss": 0.5589,
"nll_loss": 0.5211626291275024,
"rewards/accuracies": 0.5874999761581421,
"rewards/chosen": -0.044403962790966034,
"rewards/margins": 0.005955878179520369,
"rewards/rejected": -0.05035984516143799,
"step": 220
},
{
"epoch": 0.2412165705296277,
"grad_norm": 2.0148191421861705,
"learning_rate": 1.3187609467915744e-05,
"log_odds_chosen": 0.2717307209968567,
"log_odds_ratio": -0.6763201951980591,
"logits/chosen": -2.829516887664795,
"logits/rejected": -2.842909574508667,
"logps/chosen": -0.9367680549621582,
"logps/rejected": -1.1125657558441162,
"loss": 0.5701,
"nll_loss": 0.5263533592224121,
"rewards/accuracies": 0.5625,
"rewards/chosen": -0.04683841019868851,
"rewards/margins": 0.008789879269897938,
"rewards/rejected": -0.05562828853726387,
"step": 230
},
{
"epoch": 0.2517042475091767,
"grad_norm": 2.286828850039024,
"learning_rate": 1.2909944487358057e-05,
"log_odds_chosen": 0.2564060091972351,
"log_odds_ratio": -0.651031494140625,
"logits/chosen": -2.979280471801758,
"logits/rejected": -3.0063037872314453,
"logps/chosen": -0.9010913968086243,
"logps/rejected": -1.065353512763977,
"loss": 0.5799,
"nll_loss": 0.5546143054962158,
"rewards/accuracies": 0.581250011920929,
"rewards/chosen": -0.04505457356572151,
"rewards/margins": 0.008213100023567677,
"rewards/rejected": -0.053267668932676315,
"step": 240
},
{
"epoch": 0.26219192448872575,
"grad_norm": 3.959216899336302,
"learning_rate": 1.2649110640673518e-05,
"log_odds_chosen": 0.2661912143230438,
"log_odds_ratio": -0.6746715307235718,
"logits/chosen": -2.9726908206939697,
"logits/rejected": -2.974113941192627,
"logps/chosen": -0.8829942941665649,
"logps/rejected": -1.0264866352081299,
"loss": 0.5502,
"nll_loss": 0.5201153755187988,
"rewards/accuracies": 0.5874999761581421,
"rewards/chosen": -0.04414971172809601,
"rewards/margins": 0.007174622267484665,
"rewards/rejected": -0.05132433772087097,
"step": 250
},
{
"epoch": 0.2726796014682748,
"grad_norm": 2.2699181039817,
"learning_rate": 1.2403473458920845e-05,
"log_odds_chosen": 0.2342940866947174,
"log_odds_ratio": -0.6783974766731262,
"logits/chosen": -2.9759726524353027,
"logits/rejected": -2.9923360347747803,
"logps/chosen": -0.9042210578918457,
"logps/rejected": -1.0481539964675903,
"loss": 0.5304,
"nll_loss": 0.45657747983932495,
"rewards/accuracies": 0.6312500238418579,
"rewards/chosen": -0.0452110581099987,
"rewards/margins": 0.007196647580713034,
"rewards/rejected": -0.052407700568437576,
"step": 260
},
{
"epoch": 0.2831672784478238,
"grad_norm": 2.380998150273162,
"learning_rate": 1.2171612389003691e-05,
"log_odds_chosen": 0.17961958050727844,
"log_odds_ratio": -0.6983593702316284,
"logits/chosen": -2.938765525817871,
"logits/rejected": -2.965757369995117,
"logps/chosen": -0.9548166990280151,
"logps/rejected": -1.0895111560821533,
"loss": 0.5673,
"nll_loss": 0.5430372357368469,
"rewards/accuracies": 0.4937500059604645,
"rewards/chosen": -0.04774082824587822,
"rewards/margins": 0.006734730210155249,
"rewards/rejected": -0.0544755645096302,
"step": 270
},
{
"epoch": 0.29365495542737285,
"grad_norm": 2.0870887262121323,
"learning_rate": 1.1952286093343936e-05,
"log_odds_chosen": 0.2291949987411499,
"log_odds_ratio": -0.6750219464302063,
"logits/chosen": -2.928527355194092,
"logits/rejected": -2.9543163776397705,
"logps/chosen": -0.9355181455612183,
"logps/rejected": -1.0729036331176758,
"loss": 0.5434,
"nll_loss": 0.47713321447372437,
"rewards/accuracies": 0.65625,
"rewards/chosen": -0.04677591472864151,
"rewards/margins": 0.006869266740977764,
"rewards/rejected": -0.05364518240094185,
"step": 280
},
{
"epoch": 0.30414263240692185,
"grad_norm": 2.661552133228645,
"learning_rate": 1.1744404390294071e-05,
"log_odds_chosen": 0.36491650342941284,
"log_odds_ratio": -0.620793879032135,
"logits/chosen": -2.880122661590576,
"logits/rejected": -2.8935391902923584,
"logps/chosen": -0.836012065410614,
"logps/rejected": -1.05286705493927,
"loss": 0.5596,
"nll_loss": 0.4885989725589752,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.0418006032705307,
"rewards/margins": 0.010842744261026382,
"rewards/rejected": -0.05264334753155708,
"step": 290
},
{
"epoch": 0.3146303093864709,
"grad_norm": 3.127285518362044,
"learning_rate": 1.1547005383792517e-05,
"log_odds_chosen": 0.255328893661499,
"log_odds_ratio": -0.6939107179641724,
"logits/chosen": -2.9603378772735596,
"logits/rejected": -2.992128372192383,
"logps/chosen": -0.8731514811515808,
"logps/rejected": -1.0526010990142822,
"loss": 0.5835,
"nll_loss": 0.5112031102180481,
"rewards/accuracies": 0.5625,
"rewards/chosen": -0.0436575748026371,
"rewards/margins": 0.008972481824457645,
"rewards/rejected": -0.052630048245191574,
"step": 300
},
{
"epoch": 0.3251179863660199,
"grad_norm": 2.013637214040506,
"learning_rate": 1.1359236684941297e-05,
"log_odds_chosen": 0.21040907502174377,
"log_odds_ratio": -0.688109278678894,
"logits/chosen": -2.9860305786132812,
"logits/rejected": -2.9820261001586914,
"logps/chosen": -0.9089478254318237,
"logps/rejected": -1.0382112264633179,
"loss": 0.585,
"nll_loss": 0.5399721264839172,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.045447397977113724,
"rewards/margins": 0.006463165394961834,
"rewards/rejected": -0.051910560578107834,
"step": 310
},
{
"epoch": 0.33560566334556896,
"grad_norm": 2.1577553752792995,
"learning_rate": 1.118033988749895e-05,
"log_odds_chosen": 0.27985960245132446,
"log_odds_ratio": -0.6601210832595825,
"logits/chosen": -3.0387003421783447,
"logits/rejected": -3.0464096069335938,
"logps/chosen": -0.9086373448371887,
"logps/rejected": -1.0836986303329468,
"loss": 0.5243,
"nll_loss": 0.4922841191291809,
"rewards/accuracies": 0.550000011920929,
"rewards/chosen": -0.045431867241859436,
"rewards/margins": 0.008753069676458836,
"rewards/rejected": -0.0541849359869957,
"step": 320
},
{
"epoch": 0.34609334032511796,
"grad_norm": 2.422690319169778,
"learning_rate": 1.1009637651263608e-05,
"log_odds_chosen": 0.28255337476730347,
"log_odds_ratio": -0.6909259557723999,
"logits/chosen": -2.950887441635132,
"logits/rejected": -2.9948947429656982,
"logps/chosen": -0.9054603576660156,
"logps/rejected": -1.0888211727142334,
"loss": 0.5544,
"nll_loss": 0.5376341342926025,
"rewards/accuracies": 0.581250011920929,
"rewards/chosen": -0.04527302458882332,
"rewards/margins": 0.009168041869997978,
"rewards/rejected": -0.05444106459617615,
"step": 330
},
{
"epoch": 0.356581017304667,
"grad_norm": 2.2975046406882798,
"learning_rate": 1.0846522890932809e-05,
"log_odds_chosen": 0.2153971642255783,
"log_odds_ratio": -0.6926898956298828,
"logits/chosen": -2.9686572551727295,
"logits/rejected": -3.0199432373046875,
"logps/chosen": -0.8590608835220337,
"logps/rejected": -1.00636887550354,
"loss": 0.5708,
"nll_loss": 0.5127817392349243,
"rewards/accuracies": 0.543749988079071,
"rewards/chosen": -0.042953044176101685,
"rewards/margins": 0.007365405559539795,
"rewards/rejected": -0.05031844973564148,
"step": 340
},
{
"epoch": 0.36706869428421607,
"grad_norm": 2.135727653321979,
"learning_rate": 1.0690449676496977e-05,
"log_odds_chosen": 0.2665565609931946,
"log_odds_ratio": -0.6829238533973694,
"logits/chosen": -3.044860363006592,
"logits/rejected": -3.0616378784179688,
"logps/chosen": -0.8791500329971313,
"logps/rejected": -1.0402672290802002,
"loss": 0.5495,
"nll_loss": 0.5228344202041626,
"rewards/accuracies": 0.574999988079071,
"rewards/chosen": -0.04395749792456627,
"rewards/margins": 0.00805586390197277,
"rewards/rejected": -0.05201335996389389,
"step": 350
},
{
"epoch": 0.37755637126376507,
"grad_norm": 3.150177435714442,
"learning_rate": 1.0540925533894598e-05,
"log_odds_chosen": 0.4033277928829193,
"log_odds_ratio": -0.602225124835968,
"logits/chosen": -2.9472672939300537,
"logits/rejected": -2.975858211517334,
"logps/chosen": -0.8669608235359192,
"logps/rejected": -1.110353708267212,
"loss": 0.5494,
"nll_loss": 0.5087054371833801,
"rewards/accuracies": 0.625,
"rewards/chosen": -0.0433480478823185,
"rewards/margins": 0.01216964516788721,
"rewards/rejected": -0.05551769211888313,
"step": 360
},
{
"epoch": 0.3880440482433141,
"grad_norm": 2.130197231019511,
"learning_rate": 1.0397504898200728e-05,
"log_odds_chosen": 0.3966829478740692,
"log_odds_ratio": -0.6142522096633911,
"logits/chosen": -3.0528526306152344,
"logits/rejected": -3.0623490810394287,
"logps/chosen": -0.8640265464782715,
"logps/rejected": -1.1243717670440674,
"loss": 0.5232,
"nll_loss": 0.5101068615913391,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.043201327323913574,
"rewards/margins": 0.013017257675528526,
"rewards/rejected": -0.05621858313679695,
"step": 370
},
{
"epoch": 0.3985317252228631,
"grad_norm": 2.415549044992692,
"learning_rate": 1.0259783520851543e-05,
"log_odds_chosen": 0.46208301186561584,
"log_odds_ratio": -0.5873923301696777,
"logits/chosen": -3.055903196334839,
"logits/rejected": -3.089763879776001,
"logps/chosen": -0.8685981035232544,
"logps/rejected": -1.1247217655181885,
"loss": 0.5376,
"nll_loss": 0.5167646408081055,
"rewards/accuracies": 0.668749988079071,
"rewards/chosen": -0.0434299036860466,
"rewards/margins": 0.01280617993324995,
"rewards/rejected": -0.056236088275909424,
"step": 380
},
{
"epoch": 0.4090194022024122,
"grad_norm": 2.4197618087673036,
"learning_rate": 1.0127393670836667e-05,
"log_odds_chosen": 0.08936772495508194,
"log_odds_ratio": -0.7186132073402405,
"logits/chosen": -2.998857021331787,
"logits/rejected": -3.021352529525757,
"logps/chosen": -0.9128287434577942,
"logps/rejected": -0.9754525423049927,
"loss": 0.5571,
"nll_loss": 0.5319759845733643,
"rewards/accuracies": 0.5625,
"rewards/chosen": -0.04564143717288971,
"rewards/margins": 0.0031311833299696445,
"rewards/rejected": -0.048772621899843216,
"step": 390
},
{
"epoch": 0.4195070791819612,
"grad_norm": 2.0748995530757424,
"learning_rate": 1e-05,
"log_odds_chosen": 0.23965713381767273,
"log_odds_ratio": -0.6899853348731995,
"logits/chosen": -2.883575201034546,
"logits/rejected": -2.908125400543213,
"logps/chosen": -0.9490350484848022,
"logps/rejected": -1.1106139421463013,
"loss": 0.5725,
"nll_loss": 0.5262094736099243,
"rewards/accuracies": 0.53125,
"rewards/chosen": -0.04745175316929817,
"rewards/margins": 0.00807894580066204,
"rewards/rejected": -0.05553068965673447,
"step": 400
},
{
"epoch": 0.4299947561615102,
"grad_norm": 2.0498490112152026,
"learning_rate": 9.877295966495898e-06,
"log_odds_chosen": 0.14244404435157776,
"log_odds_ratio": -0.7278560996055603,
"logits/chosen": -2.988100051879883,
"logits/rejected": -2.9914164543151855,
"logps/chosen": -0.8709594011306763,
"logps/rejected": -0.9773006439208984,
"loss": 0.5455,
"nll_loss": 0.4832683503627777,
"rewards/accuracies": 0.5687500238418579,
"rewards/chosen": -0.04354798048734665,
"rewards/margins": 0.0053170593455433846,
"rewards/rejected": -0.04886503517627716,
"step": 410
},
{
"epoch": 0.4404824331410593,
"grad_norm": 1.9311064341389872,
"learning_rate": 9.759000729485331e-06,
"log_odds_chosen": 0.30063071846961975,
"log_odds_ratio": -0.643203854560852,
"logits/chosen": -2.9488558769226074,
"logits/rejected": -2.9841551780700684,
"logps/chosen": -0.8707404136657715,
"logps/rejected": -1.0532442331314087,
"loss": 0.5355,
"nll_loss": 0.474843829870224,
"rewards/accuracies": 0.612500011920929,
"rewards/chosen": -0.04353701323270798,
"rewards/margins": 0.009125196374952793,
"rewards/rejected": -0.05266221612691879,
"step": 420
},
{
"epoch": 0.4509701101206083,
"grad_norm": 2.119895291758326,
"learning_rate": 9.644856443408244e-06,
"log_odds_chosen": 0.2837393879890442,
"log_odds_ratio": -0.6551750898361206,
"logits/chosen": -2.9840757846832275,
"logits/rejected": -2.9921929836273193,
"logps/chosen": -0.8468173146247864,
"logps/rejected": -1.0135347843170166,
"loss": 0.5557,
"nll_loss": 0.5443450212478638,
"rewards/accuracies": 0.6312500238418579,
"rewards/chosen": -0.04234086349606514,
"rewards/margins": 0.00833587534725666,
"rewards/rejected": -0.05067674070596695,
"step": 430
},
{
"epoch": 0.46145778710015734,
"grad_norm": 2.095435518308805,
"learning_rate": 9.534625892455923e-06,
"log_odds_chosen": 0.2355252504348755,
"log_odds_ratio": -0.6598283648490906,
"logits/chosen": -3.0252740383148193,
"logits/rejected": -3.045849323272705,
"logps/chosen": -0.8709392547607422,
"logps/rejected": -1.0179613828659058,
"loss": 0.5508,
"nll_loss": 0.5189236998558044,
"rewards/accuracies": 0.5874999761581421,
"rewards/chosen": -0.04354696720838547,
"rewards/margins": 0.0073511130176484585,
"rewards/rejected": -0.050898075103759766,
"step": 440
},
{
"epoch": 0.47194546407970633,
"grad_norm": 1.9017756846669818,
"learning_rate": 9.428090415820635e-06,
"log_odds_chosen": 0.34075412154197693,
"log_odds_ratio": -0.6583858728408813,
"logits/chosen": -3.0218703746795654,
"logits/rejected": -3.0481696128845215,
"logps/chosen": -0.8293315768241882,
"logps/rejected": -1.047191858291626,
"loss": 0.5286,
"nll_loss": 0.4964592456817627,
"rewards/accuracies": 0.6312500238418579,
"rewards/chosen": -0.04146658256649971,
"rewards/margins": 0.010893006809055805,
"rewards/rejected": -0.05235959216952324,
"step": 450
},
{
"epoch": 0.4824331410592554,
"grad_norm": 2.079766146123277,
"learning_rate": 9.325048082403139e-06,
"log_odds_chosen": 0.16855968534946442,
"log_odds_ratio": -0.711928129196167,
"logits/chosen": -3.0086510181427,
"logits/rejected": -3.0489156246185303,
"logps/chosen": -0.9442957043647766,
"logps/rejected": -1.072997808456421,
"loss": 0.5326,
"nll_loss": 0.5338221788406372,
"rewards/accuracies": 0.543749988079071,
"rewards/chosen": -0.04721478372812271,
"rewards/margins": 0.00643510278314352,
"rewards/rejected": -0.05364988371729851,
"step": 460
},
{
"epoch": 0.4929208180388044,
"grad_norm": 2.4868491558153085,
"learning_rate": 9.225312080288851e-06,
"log_odds_chosen": 0.23586861789226532,
"log_odds_ratio": -0.6902174949645996,
"logits/chosen": -2.986264705657959,
"logits/rejected": -3.0127644538879395,
"logps/chosen": -0.8882457613945007,
"logps/rejected": -1.034985899925232,
"loss": 0.5413,
"nll_loss": 0.5090312361717224,
"rewards/accuracies": 0.5562499761581421,
"rewards/chosen": -0.044412292540073395,
"rewards/margins": 0.007337009999901056,
"rewards/rejected": -0.051749296486377716,
"step": 470
},
{
"epoch": 0.5034084950183534,
"grad_norm": 2.0043501739666882,
"learning_rate": 9.12870929175277e-06,
"log_odds_chosen": 0.17604230344295502,
"log_odds_ratio": -0.707550048828125,
"logits/chosen": -3.088604211807251,
"logits/rejected": -3.12184476852417,
"logps/chosen": -0.8456010818481445,
"logps/rejected": -0.9586717486381531,
"loss": 0.5178,
"nll_loss": 0.5126105546951294,
"rewards/accuracies": 0.5625,
"rewards/chosen": -0.042280055582523346,
"rewards/margins": 0.005653535481542349,
"rewards/rejected": -0.047933585941791534,
"step": 480
},
{
"epoch": 0.5138961719979025,
"grad_norm": 1.9415978406566505,
"learning_rate": 9.035079029052514e-06,
"log_odds_chosen": 0.22476902604103088,
"log_odds_ratio": -0.6716736555099487,
"logits/chosen": -3.003417491912842,
"logits/rejected": -3.0048608779907227,
"logps/chosen": -0.9196673631668091,
"logps/rejected": -1.0358223915100098,
"loss": 0.5397,
"nll_loss": 0.5024985671043396,
"rewards/accuracies": 0.5874999761581421,
"rewards/chosen": -0.045983362942934036,
"rewards/margins": 0.005807754583656788,
"rewards/rejected": -0.051791124045848846,
"step": 490
},
{
"epoch": 0.5243838489774515,
"grad_norm": 2.2353701695425423,
"learning_rate": 8.94427190999916e-06,
"log_odds_chosen": 0.20684054493904114,
"log_odds_ratio": -0.698712944984436,
"logits/chosen": -3.0111751556396484,
"logits/rejected": -3.0036330223083496,
"logps/chosen": -0.8826943635940552,
"logps/rejected": -1.0074814558029175,
"loss": 0.548,
"nll_loss": 0.5235316157341003,
"rewards/accuracies": 0.625,
"rewards/chosen": -0.0441347137093544,
"rewards/margins": 0.006239361595362425,
"rewards/rejected": -0.050374072045087814,
"step": 500
},
{
"epoch": 0.5348715259570005,
"grad_norm": 1.742537477144132,
"learning_rate": 8.856148855400955e-06,
"log_odds_chosen": 0.3066679835319519,
"log_odds_ratio": -0.6453306674957275,
"logits/chosen": -2.9636032581329346,
"logits/rejected": -2.97407865524292,
"logps/chosen": -0.8404191136360168,
"logps/rejected": -1.0267155170440674,
"loss": 0.5264,
"nll_loss": 0.5354185104370117,
"rewards/accuracies": 0.606249988079071,
"rewards/chosen": -0.04202095791697502,
"rewards/margins": 0.009314822033047676,
"rewards/rejected": -0.05133577436208725,
"step": 510
},
{
"epoch": 0.5453592029365496,
"grad_norm": 1.6799388590726438,
"learning_rate": 8.770580193070294e-06,
"log_odds_chosen": 0.24468369781970978,
"log_odds_ratio": -0.6710330247879028,
"logits/chosen": -2.959213972091675,
"logits/rejected": -2.966728687286377,
"logps/chosen": -0.9035038948059082,
"logps/rejected": -1.0690029859542847,
"loss": 0.5366,
"nll_loss": 0.47406935691833496,
"rewards/accuracies": 0.606249988079071,
"rewards/chosen": -0.04517520219087601,
"rewards/margins": 0.008274954743683338,
"rewards/rejected": -0.053450148552656174,
"step": 520
},
{
"epoch": 0.5558468799160986,
"grad_norm": 1.8707354612150964,
"learning_rate": 8.687444855261389e-06,
"log_odds_chosen": 0.4215427339076996,
"log_odds_ratio": -0.6489927172660828,
"logits/chosen": -3.0756938457489014,
"logits/rejected": -3.0923542976379395,
"logps/chosen": -0.8253329992294312,
"logps/rejected": -1.1108949184417725,
"loss": 0.5365,
"nll_loss": 0.45042163133621216,
"rewards/accuracies": 0.53125,
"rewards/chosen": -0.04126664996147156,
"rewards/margins": 0.014278100803494453,
"rewards/rejected": -0.05554475262761116,
"step": 530
},
{
"epoch": 0.5663345568956476,
"grad_norm": 1.922705947748225,
"learning_rate": 8.606629658238705e-06,
"log_odds_chosen": 0.1879667341709137,
"log_odds_ratio": -0.6903280019760132,
"logits/chosen": -2.975130796432495,
"logits/rejected": -3.0028696060180664,
"logps/chosen": -0.8695458173751831,
"logps/rejected": -0.9805169105529785,
"loss": 0.5535,
"nll_loss": 0.5275255441665649,
"rewards/accuracies": 0.574999988079071,
"rewards/chosen": -0.04347729682922363,
"rewards/margins": 0.005548550747334957,
"rewards/rejected": -0.049025844782590866,
"step": 540
},
{
"epoch": 0.5768222338751966,
"grad_norm": 1.9089385183272836,
"learning_rate": 8.528028654224417e-06,
"log_odds_chosen": 0.42722567915916443,
"log_odds_ratio": -0.6043616533279419,
"logits/chosen": -2.9973807334899902,
"logits/rejected": -3.0049965381622314,
"logps/chosen": -0.8592002987861633,
"logps/rejected": -1.1192405223846436,
"loss": 0.537,
"nll_loss": 0.5372708439826965,
"rewards/accuracies": 0.6312500238418579,
"rewards/chosen": -0.042960021644830704,
"rewards/margins": 0.013002010062336922,
"rewards/rejected": -0.05596202611923218,
"step": 550
},
{
"epoch": 0.5873099108547457,
"grad_norm": 1.9519454661958895,
"learning_rate": 8.451542547285167e-06,
"log_odds_chosen": 0.23686861991882324,
"log_odds_ratio": -0.679013192653656,
"logits/chosen": -3.0309016704559326,
"logits/rejected": -3.0620574951171875,
"logps/chosen": -0.8845365643501282,
"logps/rejected": -1.0314432382583618,
"loss": 0.5215,
"nll_loss": 0.5018130540847778,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.04422682151198387,
"rewards/margins": 0.0073453388176858425,
"rewards/rejected": -0.05157216265797615,
"step": 560
},
{
"epoch": 0.5977975878342947,
"grad_norm": 1.902474576616517,
"learning_rate": 8.37707816583391e-06,
"log_odds_chosen": 0.157462477684021,
"log_odds_ratio": -0.7165660858154297,
"logits/chosen": -2.971592903137207,
"logits/rejected": -2.9932913780212402,
"logps/chosen": -0.8898121118545532,
"logps/rejected": -0.9948716163635254,
"loss": 0.5041,
"nll_loss": 0.5276492834091187,
"rewards/accuracies": 0.5,
"rewards/chosen": -0.044490598142147064,
"rewards/margins": 0.005252980627119541,
"rewards/rejected": -0.04974358528852463,
"step": 570
},
{
"epoch": 0.6082852648138437,
"grad_norm": 1.9526588876095308,
"learning_rate": 8.304547985373997e-06,
"log_odds_chosen": 0.27767136693000793,
"log_odds_ratio": -0.6578360199928284,
"logits/chosen": -3.0485613346099854,
"logits/rejected": -3.061281204223633,
"logps/chosen": -0.8733240962028503,
"logps/rejected": -1.0594861507415771,
"loss": 0.5456,
"nll_loss": 0.48286086320877075,
"rewards/accuracies": 0.5687500238418579,
"rewards/chosen": -0.043666206300258636,
"rewards/margins": 0.009308096952736378,
"rewards/rejected": -0.05297430604696274,
"step": 580
},
{
"epoch": 0.6187729417933928,
"grad_norm": 1.963515177379308,
"learning_rate": 8.233869695926184e-06,
"log_odds_chosen": 0.32016056776046753,
"log_odds_ratio": -0.6649240255355835,
"logits/chosen": -3.0834898948669434,
"logits/rejected": -3.123967409133911,
"logps/chosen": -0.8281318545341492,
"logps/rejected": -1.021436095237732,
"loss": 0.5124,
"nll_loss": 0.5498961210250854,
"rewards/accuracies": 0.612500011920929,
"rewards/chosen": -0.04140659421682358,
"rewards/margins": 0.009665210731327534,
"rewards/rejected": -0.05107180029153824,
"step": 590
},
{
"epoch": 0.6292606187729418,
"grad_norm": 2.1416673571833584,
"learning_rate": 8.164965809277262e-06,
"log_odds_chosen": 0.3141978085041046,
"log_odds_ratio": -0.6486893892288208,
"logits/chosen": -3.1147074699401855,
"logits/rejected": -3.11454176902771,
"logps/chosen": -0.8215556144714355,
"logps/rejected": -1.009476661682129,
"loss": 0.5144,
"nll_loss": 0.4836875796318054,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.04107777774333954,
"rewards/margins": 0.009396053850650787,
"rewards/rejected": -0.05047383904457092,
"step": 600
},
{
"epoch": 0.6397482957524908,
"grad_norm": 2.03894912155955,
"learning_rate": 8.097763301789162e-06,
"log_odds_chosen": 0.1958848237991333,
"log_odds_ratio": -0.6933802366256714,
"logits/chosen": -3.016098737716675,
"logits/rejected": -3.046642780303955,
"logps/chosen": -0.8733209371566772,
"logps/rejected": -0.9883171916007996,
"loss": 0.526,
"nll_loss": 0.4880569875240326,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.0436660535633564,
"rewards/margins": 0.005749809555709362,
"rewards/rejected": -0.049415864050388336,
"step": 610
},
{
"epoch": 0.6502359727320398,
"grad_norm": 2.068974001178546,
"learning_rate": 8.03219328902499e-06,
"log_odds_chosen": 0.17991718649864197,
"log_odds_ratio": -0.7055822610855103,
"logits/chosen": -3.045403003692627,
"logits/rejected": -3.0644798278808594,
"logps/chosen": -0.8806620836257935,
"logps/rejected": -1.0145095586776733,
"loss": 0.5295,
"nll_loss": 0.5151625275611877,
"rewards/accuracies": 0.5375000238418579,
"rewards/chosen": -0.04403311014175415,
"rewards/margins": 0.006692370865494013,
"rewards/rejected": -0.05072547867894173,
"step": 620
},
{
"epoch": 0.6607236497115889,
"grad_norm": 1.9705491215328443,
"learning_rate": 7.968190728895958e-06,
"log_odds_chosen": 0.23948292434215546,
"log_odds_ratio": -0.6947344541549683,
"logits/chosen": -3.016519546508789,
"logits/rejected": -3.042133331298828,
"logps/chosen": -0.8557758331298828,
"logps/rejected": -1.0029237270355225,
"loss": 0.5331,
"nll_loss": 0.5245988368988037,
"rewards/accuracies": 0.53125,
"rewards/chosen": -0.0427887924015522,
"rewards/margins": 0.007357400842010975,
"rewards/rejected": -0.0501461923122406,
"step": 630
},
{
"epoch": 0.6712113266911379,
"grad_norm": 2.664256522681278,
"learning_rate": 7.905694150420949e-06,
"log_odds_chosen": 0.3717094659805298,
"log_odds_ratio": -0.6480633020401001,
"logits/chosen": -3.0543761253356934,
"logits/rejected": -3.0751733779907227,
"logps/chosen": -0.8645519018173218,
"logps/rejected": -1.102386713027954,
"loss": 0.5149,
"nll_loss": 0.46133953332901,
"rewards/accuracies": 0.574999988079071,
"rewards/chosen": -0.04322759807109833,
"rewards/margins": 0.011891739442944527,
"rewards/rejected": -0.0551193431019783,
"step": 640
},
{
"epoch": 0.6816990036706869,
"grad_norm": 1.878621524799117,
"learning_rate": 7.844645405527363e-06,
"log_odds_chosen": 0.1861819326877594,
"log_odds_ratio": -0.7022497057914734,
"logits/chosen": -3.0863146781921387,
"logits/rejected": -3.113098621368408,
"logps/chosen": -0.8403372764587402,
"logps/rejected": -0.9548438191413879,
"loss": 0.5336,
"nll_loss": 0.5122831463813782,
"rewards/accuracies": 0.59375,
"rewards/chosen": -0.04201686754822731,
"rewards/margins": 0.0057253288105130196,
"rewards/rejected": -0.047742195427417755,
"step": 650
},
{
"epoch": 0.6921866806502359,
"grad_norm": 1.8977100039056058,
"learning_rate": 7.78498944161523e-06,
"log_odds_chosen": 0.2854728400707245,
"log_odds_ratio": -0.6552462577819824,
"logits/chosen": -3.052263021469116,
"logits/rejected": -3.0898962020874023,
"logps/chosen": -0.8826674222946167,
"logps/rejected": -1.0711818933486938,
"loss": 0.5304,
"nll_loss": 0.4874996542930603,
"rewards/accuracies": 0.574999988079071,
"rewards/chosen": -0.044133372604846954,
"rewards/margins": 0.009425725787878036,
"rewards/rejected": -0.05355909466743469,
"step": 660
},
{
"epoch": 0.702674357629785,
"grad_norm": 1.8195731091765575,
"learning_rate": 7.726674092862559e-06,
"log_odds_chosen": 0.4364054203033447,
"log_odds_ratio": -0.6321254968643188,
"logits/chosen": -2.9931445121765137,
"logits/rejected": -3.025317907333374,
"logps/chosen": -0.8416171073913574,
"logps/rejected": -1.1292223930358887,
"loss": 0.5237,
"nll_loss": 0.46936100721359253,
"rewards/accuracies": 0.643750011920929,
"rewards/chosen": -0.04208085685968399,
"rewards/margins": 0.014380265958607197,
"rewards/rejected": -0.05646112561225891,
"step": 670
},
{
"epoch": 0.713162034609334,
"grad_norm": 2.0599075037830192,
"learning_rate": 7.669649888473705e-06,
"log_odds_chosen": 0.31395241618156433,
"log_odds_ratio": -0.650139570236206,
"logits/chosen": -2.9855525493621826,
"logits/rejected": -2.9897267818450928,
"logps/chosen": -0.8750125169754028,
"logps/rejected": -1.0669299364089966,
"loss": 0.5075,
"nll_loss": 0.4943002760410309,
"rewards/accuracies": 0.6312500238418579,
"rewards/chosen": -0.04375062882900238,
"rewards/margins": 0.009595867246389389,
"rewards/rejected": -0.05334649235010147,
"step": 680
},
{
"epoch": 0.723649711588883,
"grad_norm": 1.8347271674417223,
"learning_rate": 7.61386987626881e-06,
"log_odds_chosen": 0.18291696906089783,
"log_odds_ratio": -0.7239105701446533,
"logits/chosen": -2.97595477104187,
"logits/rejected": -2.991725444793701,
"logps/chosen": -0.8641953468322754,
"logps/rejected": -0.9991108179092407,
"loss": 0.5304,
"nll_loss": 0.5499680638313293,
"rewards/accuracies": 0.550000011920929,
"rewards/chosen": -0.04320976510643959,
"rewards/margins": 0.006745772901922464,
"rewards/rejected": -0.04995553940534592,
"step": 690
},
{
"epoch": 0.7341373885684321,
"grad_norm": 2.2852704943230915,
"learning_rate": 7.559289460184545e-06,
"log_odds_chosen": 0.3105728030204773,
"log_odds_ratio": -0.6319602727890015,
"logits/chosen": -2.985989809036255,
"logits/rejected": -3.0209579467773438,
"logps/chosen": -0.8320032358169556,
"logps/rejected": -1.0303562879562378,
"loss": 0.5296,
"nll_loss": 0.5422422885894775,
"rewards/accuracies": 0.6187499761581421,
"rewards/chosen": -0.04160016402602196,
"rewards/margins": 0.009917653165757656,
"rewards/rejected": -0.05151782184839249,
"step": 700
},
{
"epoch": 0.7446250655479811,
"grad_norm": 1.9768197452256755,
"learning_rate": 7.505866250408016e-06,
"log_odds_chosen": 0.2948063015937805,
"log_odds_ratio": -0.6451742649078369,
"logits/chosen": -3.1170597076416016,
"logits/rejected": -3.136089324951172,
"logps/chosen": -0.8415013551712036,
"logps/rejected": -1.0454984903335571,
"loss": 0.5237,
"nll_loss": 0.47949719429016113,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.04207506403326988,
"rewards/margins": 0.01019985694438219,
"rewards/rejected": -0.0522749237716198,
"step": 710
},
{
"epoch": 0.7551127425275301,
"grad_norm": 1.905599119477425,
"learning_rate": 7.4535599249993e-06,
"log_odds_chosen": 0.40306347608566284,
"log_odds_ratio": -0.6352882385253906,
"logits/chosen": -3.064483642578125,
"logits/rejected": -3.087808847427368,
"logps/chosen": -0.7972971200942993,
"logps/rejected": -1.046507477760315,
"loss": 0.5304,
"nll_loss": 0.4636651873588562,
"rewards/accuracies": 0.625,
"rewards/chosen": -0.039864856749773026,
"rewards/margins": 0.012460513040423393,
"rewards/rejected": -0.05232536792755127,
"step": 720
},
{
"epoch": 0.7656004195070791,
"grad_norm": 2.19124615484763,
"learning_rate": 7.402332101976053e-06,
"log_odds_chosen": 0.12367966026067734,
"log_odds_ratio": -0.7226089239120483,
"logits/chosen": -3.0835583209991455,
"logits/rejected": -3.0826332569122314,
"logps/chosen": -0.8365408778190613,
"logps/rejected": -0.9029885530471802,
"loss": 0.5374,
"nll_loss": 0.5031268000602722,
"rewards/accuracies": 0.5375000238418579,
"rewards/chosen": -0.04182704538106918,
"rewards/margins": 0.0033223754726350307,
"rewards/rejected": -0.04514942690730095,
"step": 730
},
{
"epoch": 0.7760880964866282,
"grad_norm": 2.0835998895674837,
"learning_rate": 7.352146220938079e-06,
"log_odds_chosen": 0.33691075444221497,
"log_odds_ratio": -0.6264201402664185,
"logits/chosen": -3.1278512477874756,
"logits/rejected": -3.139995574951172,
"logps/chosen": -0.8067742586135864,
"logps/rejected": -1.0221493244171143,
"loss": 0.5312,
"nll_loss": 0.4790155291557312,
"rewards/accuracies": 0.668749988079071,
"rewards/chosen": -0.040338706225156784,
"rewards/margins": 0.010768752545118332,
"rewards/rejected": -0.051107458770275116,
"step": 740
},
{
"epoch": 0.7865757734661772,
"grad_norm": 1.9667031119071154,
"learning_rate": 7.3029674334022146e-06,
"log_odds_chosen": 0.23670358955860138,
"log_odds_ratio": -0.6752098202705383,
"logits/chosen": -3.1056113243103027,
"logits/rejected": -3.1298460960388184,
"logps/chosen": -0.8614869117736816,
"logps/rejected": -0.9949930310249329,
"loss": 0.5426,
"nll_loss": 0.4975660443305969,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.0430743470788002,
"rewards/margins": 0.006675302051007748,
"rewards/rejected": -0.049749650061130524,
"step": 750
},
{
"epoch": 0.7970634504457262,
"grad_norm": 1.8638714551633075,
"learning_rate": 7.254762501100117e-06,
"log_odds_chosen": 0.2394195795059204,
"log_odds_ratio": -0.6686865091323853,
"logits/chosen": -3.092322826385498,
"logits/rejected": -3.0998446941375732,
"logps/chosen": -0.8189753293991089,
"logps/rejected": -0.9735254049301147,
"loss": 0.5115,
"nll_loss": 0.4049908220767975,
"rewards/accuracies": 0.5687500238418579,
"rewards/chosen": -0.0409487709403038,
"rewards/margins": 0.007727508433163166,
"rewards/rejected": -0.048676274716854095,
"step": 760
},
{
"epoch": 0.8075511274252754,
"grad_norm": 2.098087236150393,
"learning_rate": 7.207499701564472e-06,
"log_odds_chosen": 0.21572642028331757,
"log_odds_ratio": -0.7029857635498047,
"logits/chosen": -3.0059127807617188,
"logits/rejected": -3.0258781909942627,
"logps/chosen": -0.8941653370857239,
"logps/rejected": -1.0438942909240723,
"loss": 0.5343,
"nll_loss": 0.5011810064315796,
"rewards/accuracies": 0.48750001192092896,
"rewards/chosen": -0.044708263128995895,
"rewards/margins": 0.007486448623239994,
"rewards/rejected": -0.05219471454620361,
"step": 770
},
{
"epoch": 0.8180388044048243,
"grad_norm": 1.908201970451478,
"learning_rate": 7.1611487403943295e-06,
"log_odds_chosen": 0.22588184475898743,
"log_odds_ratio": -0.6703106164932251,
"logits/chosen": -3.0057101249694824,
"logits/rejected": -3.0319108963012695,
"logps/chosen": -0.8802768588066101,
"logps/rejected": -0.997613787651062,
"loss": 0.5466,
"nll_loss": 0.5490036606788635,
"rewards/accuracies": 0.612500011920929,
"rewards/chosen": -0.0440138503909111,
"rewards/margins": 0.005866837687790394,
"rewards/rejected": -0.04988069087266922,
"step": 780
},
{
"epoch": 0.8285264813843733,
"grad_norm": 1.8452821315553456,
"learning_rate": 7.115680669648201e-06,
"log_odds_chosen": 0.32251420617103577,
"log_odds_ratio": -0.6489396691322327,
"logits/chosen": -2.991415500640869,
"logits/rejected": -3.0075478553771973,
"logps/chosen": -0.8143788576126099,
"logps/rejected": -1.0171436071395874,
"loss": 0.5052,
"nll_loss": 0.4423222541809082,
"rewards/accuracies": 0.5874999761581421,
"rewards/chosen": -0.04071894288063049,
"rewards/margins": 0.010138243436813354,
"rewards/rejected": -0.05085718631744385,
"step": 790
},
{
"epoch": 0.8390141583639223,
"grad_norm": 2.099723593564682,
"learning_rate": 7.0710678118654756e-06,
"log_odds_chosen": 0.4498319625854492,
"log_odds_ratio": -0.5986544489860535,
"logits/chosen": -2.9999208450317383,
"logits/rejected": -2.9963490962982178,
"logps/chosen": -0.782555341720581,
"logps/rejected": -1.068285584449768,
"loss": 0.5173,
"nll_loss": 0.4201901853084564,
"rewards/accuracies": 0.65625,
"rewards/chosen": -0.03912776708602905,
"rewards/margins": 0.014286505989730358,
"rewards/rejected": -0.053414274007081985,
"step": 800
},
{
"epoch": 0.8495018353434715,
"grad_norm": 1.9010573028789273,
"learning_rate": 7.027283689263066e-06,
"log_odds_chosen": 0.34422335028648376,
"log_odds_ratio": -0.6322020292282104,
"logits/chosen": -3.0011842250823975,
"logits/rejected": -2.9966137409210205,
"logps/chosen": -0.8086786270141602,
"logps/rejected": -1.0155996084213257,
"loss": 0.5132,
"nll_loss": 0.4740920066833496,
"rewards/accuracies": 0.59375,
"rewards/chosen": -0.040433935821056366,
"rewards/margins": 0.010346042923629284,
"rewards/rejected": -0.05077998712658882,
"step": 810
},
{
"epoch": 0.8599895123230205,
"grad_norm": 2.3144073315770353,
"learning_rate": 6.984302957695783e-06,
"log_odds_chosen": 0.29515784978866577,
"log_odds_ratio": -0.6521409749984741,
"logits/chosen": -2.943692445755005,
"logits/rejected": -2.9414219856262207,
"logps/chosen": -0.8414862751960754,
"logps/rejected": -1.0143965482711792,
"loss": 0.504,
"nll_loss": 0.4271189570426941,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.04207431524991989,
"rewards/margins": 0.008645516820251942,
"rewards/rejected": -0.05071982741355896,
"step": 820
},
{
"epoch": 0.8704771893025695,
"grad_norm": 2.371001107698096,
"learning_rate": 6.942101345006233e-06,
"log_odds_chosen": 0.2455742061138153,
"log_odds_ratio": -0.7013689279556274,
"logits/chosen": -2.933568239212036,
"logits/rejected": -2.977832794189453,
"logps/chosen": -0.8553229570388794,
"logps/rejected": -1.0332233905792236,
"loss": 0.5251,
"nll_loss": 0.46586036682128906,
"rewards/accuracies": 0.53125,
"rewards/chosen": -0.04276614636182785,
"rewards/margins": 0.008895025588572025,
"rewards/rejected": -0.0516611710190773,
"step": 830
},
{
"epoch": 0.8809648662821186,
"grad_norm": 1.977587507180873,
"learning_rate": 6.900655593423542e-06,
"log_odds_chosen": 0.19387319684028625,
"log_odds_ratio": -0.6939007639884949,
"logits/chosen": -2.9483094215393066,
"logits/rejected": -2.966421365737915,
"logps/chosen": -0.8696029782295227,
"logps/rejected": -1.0034617185592651,
"loss": 0.5136,
"nll_loss": 0.48451894521713257,
"rewards/accuracies": 0.543749988079071,
"rewards/chosen": -0.043480150401592255,
"rewards/margins": 0.006692938506603241,
"rewards/rejected": -0.050173092633485794,
"step": 840
},
{
"epoch": 0.8914525432616676,
"grad_norm": 2.0931872980265527,
"learning_rate": 6.859943405700353e-06,
"log_odds_chosen": 0.27469760179519653,
"log_odds_ratio": -0.6496983170509338,
"logits/chosen": -2.882544994354248,
"logits/rejected": -2.907102584838867,
"logps/chosen": -0.8309645652770996,
"logps/rejected": -0.9983605146408081,
"loss": 0.5054,
"nll_loss": 0.4892002046108246,
"rewards/accuracies": 0.59375,
"rewards/chosen": -0.0415482297539711,
"rewards/margins": 0.008369805291295052,
"rewards/rejected": -0.049918033182621,
"step": 850
},
{
"epoch": 0.9019402202412166,
"grad_norm": 1.9059523373512675,
"learning_rate": 6.819943394704736e-06,
"log_odds_chosen": 0.2372780740261078,
"log_odds_ratio": -0.6811105012893677,
"logits/chosen": -2.9579243659973145,
"logits/rejected": -2.9706907272338867,
"logps/chosen": -0.8282278180122375,
"logps/rejected": -0.982342541217804,
"loss": 0.5277,
"nll_loss": 0.4725598692893982,
"rewards/accuracies": 0.543749988079071,
"rewards/chosen": -0.0414113886654377,
"rewards/margins": 0.007705743424594402,
"rewards/rejected": -0.049117133021354675,
"step": 860
},
{
"epoch": 0.9124278972207656,
"grad_norm": 1.892543797666968,
"learning_rate": 6.780635036208105e-06,
"log_odds_chosen": 0.287548691034317,
"log_odds_ratio": -0.6644268035888672,
"logits/chosen": -3.0049710273742676,
"logits/rejected": -3.0431902408599854,
"logps/chosen": -0.8620280027389526,
"logps/rejected": -1.0551369190216064,
"loss": 0.4935,
"nll_loss": 0.4828346371650696,
"rewards/accuracies": 0.5625,
"rewards/chosen": -0.04310140386223793,
"rewards/margins": 0.009655444882810116,
"rewards/rejected": -0.05275684595108032,
"step": 870
},
{
"epoch": 0.9229155742003147,
"grad_norm": 1.6128728864363475,
"learning_rate": 6.741998624632421e-06,
"log_odds_chosen": 0.2844703197479248,
"log_odds_ratio": -0.6617631316184998,
"logits/chosen": -3.044353723526001,
"logits/rejected": -3.0480034351348877,
"logps/chosen": -0.808245837688446,
"logps/rejected": -0.990073561668396,
"loss": 0.4881,
"nll_loss": 0.43747878074645996,
"rewards/accuracies": 0.606249988079071,
"rewards/chosen": -0.0404122956097126,
"rewards/margins": 0.009091392159461975,
"rewards/rejected": -0.04950368404388428,
"step": 880
},
{
"epoch": 0.9334032511798637,
"grad_norm": 2.329046484142618,
"learning_rate": 6.70401523153991e-06,
"log_odds_chosen": 0.32051050662994385,
"log_odds_ratio": -0.6461818218231201,
"logits/chosen": -3.0071539878845215,
"logits/rejected": -3.0232186317443848,
"logps/chosen": -0.8105939030647278,
"logps/rejected": -0.993729293346405,
"loss": 0.4935,
"nll_loss": 0.46434158086776733,
"rewards/accuracies": 0.5874999761581421,
"rewards/chosen": -0.04052969440817833,
"rewards/margins": 0.00915677472949028,
"rewards/rejected": -0.04968646913766861,
"step": 890
},
{
"epoch": 0.9438909281594127,
"grad_norm": 2.0086740635642073,
"learning_rate": 6.666666666666667e-06,
"log_odds_chosen": 0.2798821032047272,
"log_odds_ratio": -0.664302408695221,
"logits/chosen": -2.9259209632873535,
"logits/rejected": -2.9381814002990723,
"logps/chosen": -0.7818757891654968,
"logps/rejected": -0.9571603536605835,
"loss": 0.5239,
"nll_loss": 0.4661863446235657,
"rewards/accuracies": 0.581250011920929,
"rewards/chosen": -0.03909378498792648,
"rewards/margins": 0.008764232508838177,
"rewards/rejected": -0.04785802215337753,
"step": 900
},
{
"epoch": 0.9543786051389617,
"grad_norm": 2.068822950454407,
"learning_rate": 6.629935441317959e-06,
"log_odds_chosen": 0.479647159576416,
"log_odds_ratio": -0.6314842700958252,
"logits/chosen": -2.974902629852295,
"logits/rejected": -2.9787256717681885,
"logps/chosen": -0.8285977244377136,
"logps/rejected": -1.1534996032714844,
"loss": 0.5142,
"nll_loss": 0.46572408080101013,
"rewards/accuracies": 0.612500011920929,
"rewards/chosen": -0.04142988473176956,
"rewards/margins": 0.016245096921920776,
"rewards/rejected": -0.05767498165369034,
"step": 910
},
{
"epoch": 0.9648662821185108,
"grad_norm": 1.9606527520032064,
"learning_rate": 6.593804733957872e-06,
"log_odds_chosen": 0.3219223618507385,
"log_odds_ratio": -0.649006187915802,
"logits/chosen": -2.895038604736328,
"logits/rejected": -2.9138269424438477,
"logps/chosen": -0.7895429134368896,
"logps/rejected": -0.9961126446723938,
"loss": 0.4837,
"nll_loss": 0.43109196424484253,
"rewards/accuracies": 0.6187499761581421,
"rewards/chosen": -0.0394771471619606,
"rewards/margins": 0.010328484699130058,
"rewards/rejected": -0.04980562627315521,
"step": 920
},
{
"epoch": 0.9753539590980598,
"grad_norm": 2.2191050074705405,
"learning_rate": 6.55825835783953e-06,
"log_odds_chosen": 0.21952304244041443,
"log_odds_ratio": -0.6805615425109863,
"logits/chosen": -2.8973617553710938,
"logits/rejected": -2.900251865386963,
"logps/chosen": -0.8730388879776001,
"logps/rejected": -1.0255097150802612,
"loss": 0.5135,
"nll_loss": 0.5237925052642822,
"rewards/accuracies": 0.581250011920929,
"rewards/chosen": -0.04365193843841553,
"rewards/margins": 0.007623549550771713,
"rewards/rejected": -0.05127548426389694,
"step": 930
},
{
"epoch": 0.9858416360776088,
"grad_norm": 1.9816052115352747,
"learning_rate": 6.523280730534423e-06,
"log_odds_chosen": 0.2554723024368286,
"log_odds_ratio": -0.6887288689613342,
"logits/chosen": -2.93623685836792,
"logits/rejected": -2.9283607006073,
"logps/chosen": -0.7786284685134888,
"logps/rejected": -0.9273189306259155,
"loss": 0.5095,
"nll_loss": 0.4773116111755371,
"rewards/accuracies": 0.5874999761581421,
"rewards/chosen": -0.03893141821026802,
"rewards/margins": 0.007434530649334192,
"rewards/rejected": -0.046365950256586075,
"step": 940
},
{
"epoch": 0.9963293130571579,
"grad_norm": 2.074452011854083,
"learning_rate": 6.488856845230502e-06,
"log_odds_chosen": 0.2605803310871124,
"log_odds_ratio": -0.6914502382278442,
"logits/chosen": -2.9090209007263184,
"logits/rejected": -2.9163012504577637,
"logps/chosen": -0.8585780262947083,
"logps/rejected": -1.0175925493240356,
"loss": 0.5383,
"nll_loss": 0.503527045249939,
"rewards/accuracies": 0.5625,
"rewards/chosen": -0.04292890429496765,
"rewards/margins": 0.007950720377266407,
"rewards/rejected": -0.050879620015621185,
"step": 950
},
{
"epoch": 0.9994756161510225,
"step": 953,
"total_flos": 0.0,
"train_loss": 0.56347276506494,
"train_runtime": 19079.6454,
"train_samples_per_second": 3.197,
"train_steps_per_second": 0.05
}
],
"logging_steps": 10,
"max_steps": 953,
"num_input_tokens_seen": 0,
"num_train_epochs": 1,
"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": 8,
"trial_name": null,
"trial_params": null
}