qwen2.5-0.5b-expo-DPO-25-2 / trainer_state.json
hZzy's picture
Model save
485029b verified
{
"best_metric": null,
"best_model_checkpoint": null,
"epoch": 0.9995206136145733,
"eval_steps": 50,
"global_step": 695,
"is_hyper_param_search": false,
"is_local_process_zero": true,
"is_world_process_zero": true,
"log_history": [
{
"dpo_loss": 0.6931471824645996,
"epoch": 0.0014381591562799617,
"grad_norm": 19.134574294723603,
"learning_rate": 7.142857142857144e-08,
"logits": -1.475965142250061,
"logps": -86.5583724975586,
"loss": 0.6931,
"objective": 0.6931471824645996,
"ranking_simple": 0.3958333432674408,
"step": 1
},
{
"dpo_loss": 0.6926321983337402,
"epoch": 0.007190795781399808,
"grad_norm": 20.478323427973958,
"learning_rate": 3.5714285714285716e-07,
"logits": -1.5228310823440552,
"logps": -85.78047943115234,
"loss": 0.6923,
"objective": 0.6926321983337402,
"ranking_simple": 0.5625,
"step": 5
},
{
"dpo_loss": 0.6972719430923462,
"epoch": 0.014381591562799617,
"grad_norm": 18.41490700600558,
"learning_rate": 7.142857142857143e-07,
"logits": -1.5104866027832031,
"logps": -85.7982406616211,
"loss": 0.6947,
"objective": 0.6972719430923462,
"ranking_simple": 0.47083333134651184,
"step": 10
},
{
"dpo_loss": 0.6982707977294922,
"epoch": 0.021572387344199424,
"grad_norm": 18.09593626635207,
"learning_rate": 1.0714285714285714e-06,
"logits": -1.5033611059188843,
"logps": -84.19438171386719,
"loss": 0.6937,
"objective": 0.6982707977294922,
"ranking_simple": 0.5208333134651184,
"step": 15
},
{
"dpo_loss": 0.6746294498443604,
"epoch": 0.028763183125599234,
"grad_norm": 16.989375292416742,
"learning_rate": 1.4285714285714286e-06,
"logits": -1.4866502285003662,
"logps": -83.14163970947266,
"loss": 0.6849,
"objective": 0.6746294498443604,
"ranking_simple": 0.46666666865348816,
"step": 20
},
{
"dpo_loss": 0.6969509720802307,
"epoch": 0.03595397890699904,
"grad_norm": 20.67293748562133,
"learning_rate": 1.7857142857142859e-06,
"logits": -1.5258924961090088,
"logps": -85.8170166015625,
"loss": 0.6851,
"objective": 0.6969509720802307,
"ranking_simple": 0.4791666567325592,
"step": 25
},
{
"dpo_loss": 0.6758856177330017,
"epoch": 0.04314477468839885,
"grad_norm": 18.426008220098865,
"learning_rate": 2.1428571428571427e-06,
"logits": -1.4745742082595825,
"logps": -83.44979095458984,
"loss": 0.6952,
"objective": 0.6758856177330017,
"ranking_simple": 0.4625000059604645,
"step": 30
},
{
"dpo_loss": 0.6962794065475464,
"epoch": 0.050335570469798654,
"grad_norm": 17.9032967512842,
"learning_rate": 2.5e-06,
"logits": -1.4695559740066528,
"logps": -84.0098648071289,
"loss": 0.6801,
"objective": 0.6962794065475464,
"ranking_simple": 0.512499988079071,
"step": 35
},
{
"dpo_loss": 0.6491132378578186,
"epoch": 0.05752636625119847,
"grad_norm": 19.914404961281203,
"learning_rate": 2.8571428571428573e-06,
"logits": -1.5081710815429688,
"logps": -87.46798706054688,
"loss": 0.6554,
"objective": 0.6491132378578186,
"ranking_simple": 0.5291666388511658,
"step": 40
},
{
"dpo_loss": 0.6948689818382263,
"epoch": 0.06471716203259828,
"grad_norm": 18.11322171833476,
"learning_rate": 3.2142857142857147e-06,
"logits": -1.5364570617675781,
"logps": -93.01942443847656,
"loss": 0.68,
"objective": 0.6948689818382263,
"ranking_simple": 0.5458333492279053,
"step": 45
},
{
"dpo_loss": 0.6870535016059875,
"epoch": 0.07190795781399809,
"grad_norm": 16.24249291651068,
"learning_rate": 3.5714285714285718e-06,
"logits": -1.4616867303848267,
"logps": -98.69449615478516,
"loss": 0.6788,
"objective": 0.6870535016059875,
"ranking_simple": 0.5166666507720947,
"step": 50
},
{
"epoch": 0.07190795781399809,
"eval_dpo_loss": 0.6872744560241699,
"eval_logits": -1.4424751996994019,
"eval_logps": -101.08426666259766,
"eval_loss": 0.6864251494407654,
"eval_objective": 0.6872744560241699,
"eval_ranking_simple": 0.5170270204544067,
"eval_runtime": 667.7982,
"eval_samples_per_second": 16.605,
"eval_steps_per_second": 1.385,
"step": 50
},
{
"dpo_loss": 0.6663610935211182,
"epoch": 0.07909875359539789,
"grad_norm": 16.292750687919423,
"learning_rate": 3.928571428571429e-06,
"logits": -1.4286314249038696,
"logps": -99.18305206298828,
"loss": 0.653,
"objective": 0.6663610935211182,
"ranking_simple": 0.4583333432674408,
"step": 55
},
{
"dpo_loss": 0.6659743189811707,
"epoch": 0.0862895493767977,
"grad_norm": 16.962152501970213,
"learning_rate": 4.2857142857142855e-06,
"logits": -1.4745826721191406,
"logps": -93.86702728271484,
"loss": 0.6482,
"objective": 0.6659743189811707,
"ranking_simple": 0.5625,
"step": 60
},
{
"dpo_loss": 0.671344518661499,
"epoch": 0.0934803451581975,
"grad_norm": 14.901029294680688,
"learning_rate": 4.642857142857144e-06,
"logits": -1.451969027519226,
"logps": -87.68860626220703,
"loss": 0.6615,
"objective": 0.671344518661499,
"ranking_simple": 0.550000011920929,
"step": 65
},
{
"dpo_loss": 0.6434452533721924,
"epoch": 0.10067114093959731,
"grad_norm": 14.778236920418157,
"learning_rate": 5e-06,
"logits": -1.3780988454818726,
"logps": -89.031982421875,
"loss": 0.6407,
"objective": 0.6434452533721924,
"ranking_simple": 0.5833333134651184,
"step": 70
},
{
"dpo_loss": 0.6859174370765686,
"epoch": 0.10786193672099713,
"grad_norm": 28.274152674131496,
"learning_rate": 4.99921047320825e-06,
"logits": -1.4645949602127075,
"logps": -88.36033630371094,
"loss": 0.685,
"objective": 0.6859174370765686,
"ranking_simple": 0.5,
"step": 75
},
{
"dpo_loss": 0.6891925930976868,
"epoch": 0.11505273250239693,
"grad_norm": 15.764669217558497,
"learning_rate": 4.996842391515045e-06,
"logits": -1.4844069480895996,
"logps": -95.44277954101562,
"loss": 0.6754,
"objective": 0.6891925930976868,
"ranking_simple": 0.5625,
"step": 80
},
{
"dpo_loss": 0.667593777179718,
"epoch": 0.12224352828379674,
"grad_norm": 14.453520081391234,
"learning_rate": 4.992897250651535e-06,
"logits": -1.5593804121017456,
"logps": -94.26665496826172,
"loss": 0.6563,
"objective": 0.667593777179718,
"ranking_simple": 0.5166666507720947,
"step": 85
},
{
"dpo_loss": 0.6684335470199585,
"epoch": 0.12943432406519656,
"grad_norm": 13.861895733772226,
"learning_rate": 4.9873775424532515e-06,
"logits": -1.7217061519622803,
"logps": -93.05764770507812,
"loss": 0.6527,
"objective": 0.6684335470199585,
"ranking_simple": 0.5666666626930237,
"step": 90
},
{
"dpo_loss": 0.6190710663795471,
"epoch": 0.13662511984659637,
"grad_norm": 14.681765981277032,
"learning_rate": 4.980286753286196e-06,
"logits": -1.6066279411315918,
"logps": -90.85689544677734,
"loss": 0.619,
"objective": 0.6190710663795471,
"ranking_simple": 0.5375000238418579,
"step": 95
},
{
"dpo_loss": 0.6605519652366638,
"epoch": 0.14381591562799617,
"grad_norm": 15.531532836382373,
"learning_rate": 4.971629361844785e-06,
"logits": -1.7002263069152832,
"logps": -95.20433044433594,
"loss": 0.6396,
"objective": 0.6605519652366638,
"ranking_simple": 0.5625,
"step": 100
},
{
"epoch": 0.14381591562799617,
"eval_dpo_loss": 0.6830798387527466,
"eval_logits": -1.6394028663635254,
"eval_logps": -96.63552856445312,
"eval_loss": 0.6795840859413147,
"eval_objective": 0.6830798387527466,
"eval_ranking_simple": 0.5283783674240112,
"eval_runtime": 667.5921,
"eval_samples_per_second": 16.61,
"eval_steps_per_second": 1.386,
"step": 100
},
{
"dpo_loss": 0.6575779318809509,
"epoch": 0.15100671140939598,
"grad_norm": 13.9124802140197,
"learning_rate": 4.961410836323014e-06,
"logits": -1.6909289360046387,
"logps": -95.10658264160156,
"loss": 0.6472,
"objective": 0.6575779318809509,
"ranking_simple": 0.5666666626930237,
"step": 105
},
{
"dpo_loss": 0.6442916393280029,
"epoch": 0.15819750719079578,
"grad_norm": 15.449248259744447,
"learning_rate": 4.949637630960618e-06,
"logits": -1.753013253211975,
"logps": -100.01918029785156,
"loss": 0.6257,
"objective": 0.6442916393280029,
"ranking_simple": 0.574999988079071,
"step": 110
},
{
"dpo_loss": 0.5997211337089539,
"epoch": 0.1653883029721956,
"grad_norm": 15.359232778269337,
"learning_rate": 4.9363171819664434e-06,
"logits": -1.7351175546646118,
"logps": -103.54134368896484,
"loss": 0.6191,
"objective": 0.5997211337089539,
"ranking_simple": 0.5625,
"step": 115
},
{
"dpo_loss": 0.6286367774009705,
"epoch": 0.1725790987535954,
"grad_norm": 16.18007141994783,
"learning_rate": 4.921457902821578e-06,
"logits": -1.6781164407730103,
"logps": -102.82028198242188,
"loss": 0.6188,
"objective": 0.6286367774009705,
"ranking_simple": 0.49166667461395264,
"step": 120
},
{
"dpo_loss": 0.5925440192222595,
"epoch": 0.1797698945349952,
"grad_norm": 16.36914258606562,
"learning_rate": 4.905069178965215e-06,
"logits": -1.587238073348999,
"logps": -102.64461517333984,
"loss": 0.6077,
"objective": 0.5925440192222595,
"ranking_simple": 0.6499999761581421,
"step": 125
},
{
"dpo_loss": 0.5784808397293091,
"epoch": 0.186960690316395,
"grad_norm": 13.104316399935197,
"learning_rate": 4.887161361866608e-06,
"logits": -1.5238862037658691,
"logps": -98.6126708984375,
"loss": 0.5683,
"objective": 0.5784808397293091,
"ranking_simple": 0.5916666388511658,
"step": 130
},
{
"dpo_loss": 0.5823652744293213,
"epoch": 0.1941514860977948,
"grad_norm": 13.32781172011542,
"learning_rate": 4.867745762486862e-06,
"logits": -1.4705675840377808,
"logps": -96.02190399169922,
"loss": 0.6094,
"objective": 0.5823652744293213,
"ranking_simple": 0.6041666865348816,
"step": 135
},
{
"dpo_loss": 0.6114974617958069,
"epoch": 0.20134228187919462,
"grad_norm": 14.266153939024862,
"learning_rate": 4.846834644134686e-06,
"logits": -1.4452903270721436,
"logps": -95.41686248779297,
"loss": 0.643,
"objective": 0.6114974617958069,
"ranking_simple": 0.5375000238418579,
"step": 140
},
{
"dpo_loss": 0.6222321391105652,
"epoch": 0.20853307766059445,
"grad_norm": 13.527433051608822,
"learning_rate": 4.824441214720629e-06,
"logits": -1.343421220779419,
"logps": -90.63035583496094,
"loss": 0.6004,
"objective": 0.6222321391105652,
"ranking_simple": 0.5249999761581421,
"step": 145
},
{
"dpo_loss": 0.6666644811630249,
"epoch": 0.21572387344199426,
"grad_norm": 14.387730896252268,
"learning_rate": 4.800579618414677e-06,
"logits": -1.3223589658737183,
"logps": -89.82644653320312,
"loss": 0.6428,
"objective": 0.6666644811630249,
"ranking_simple": 0.5625,
"step": 150
},
{
"epoch": 0.21572387344199426,
"eval_dpo_loss": 0.6741421818733215,
"eval_logits": -1.3248826265335083,
"eval_logps": -90.46123504638672,
"eval_loss": 0.6731191277503967,
"eval_objective": 0.6741421818733215,
"eval_ranking_simple": 0.5421621799468994,
"eval_runtime": 667.0734,
"eval_samples_per_second": 16.623,
"eval_steps_per_second": 1.387,
"step": 150
},
{
"dpo_loss": 0.6579559445381165,
"epoch": 0.22291466922339406,
"grad_norm": 13.510175858510676,
"learning_rate": 4.775264926712489e-06,
"logits": -1.4420989751815796,
"logps": -87.71839904785156,
"loss": 0.6137,
"objective": 0.6579559445381165,
"ranking_simple": 0.5458333492279053,
"step": 155
},
{
"dpo_loss": 0.6222509741783142,
"epoch": 0.23010546500479387,
"grad_norm": 11.574891409598989,
"learning_rate": 4.748513128915928e-06,
"logits": -1.3550127744674683,
"logps": -82.75563049316406,
"loss": 0.6097,
"objective": 0.6222509741783142,
"ranking_simple": 0.6166666746139526,
"step": 160
},
{
"dpo_loss": 0.5967528820037842,
"epoch": 0.23729626078619367,
"grad_norm": 13.289004474817304,
"learning_rate": 4.720341122033862e-06,
"logits": -1.309799313545227,
"logps": -83.7621078491211,
"loss": 0.6238,
"objective": 0.5967528820037842,
"ranking_simple": 0.5916666388511658,
"step": 165
},
{
"dpo_loss": 0.6145275831222534,
"epoch": 0.24448705656759348,
"grad_norm": 13.95960374523932,
"learning_rate": 4.690766700109659e-06,
"logits": -1.4246125221252441,
"logps": -81.2165756225586,
"loss": 0.6133,
"objective": 0.6145275831222534,
"ranking_simple": 0.5458333492279053,
"step": 170
},
{
"dpo_loss": 0.6186246275901794,
"epoch": 0.2516778523489933,
"grad_norm": 14.582408342886906,
"learning_rate": 4.659808542982089e-06,
"logits": -1.4199490547180176,
"logps": -83.35224914550781,
"loss": 0.5772,
"objective": 0.6186246275901794,
"ranking_simple": 0.574999988079071,
"step": 175
},
{
"dpo_loss": 0.6175665259361267,
"epoch": 0.2588686481303931,
"grad_norm": 13.512566329291777,
"learning_rate": 4.62748620448673e-06,
"logits": -1.5705552101135254,
"logps": -82.73717498779297,
"loss": 0.5669,
"objective": 0.6175665259361267,
"ranking_simple": 0.5583333373069763,
"step": 180
},
{
"dpo_loss": 0.5879412293434143,
"epoch": 0.2660594439117929,
"grad_norm": 13.766981915987873,
"learning_rate": 4.593820100105355e-06,
"logits": -1.5317676067352295,
"logps": -81.35298919677734,
"loss": 0.5636,
"objective": 0.5879412293434143,
"ranking_simple": 0.6625000238418579,
"step": 185
},
{
"dpo_loss": 0.576286256313324,
"epoch": 0.27325023969319273,
"grad_norm": 15.043077064111454,
"learning_rate": 4.558831494071069e-06,
"logits": -1.59921395778656,
"logps": -86.69087219238281,
"loss": 0.5783,
"objective": 0.576286256313324,
"ranking_simple": 0.6208333373069763,
"step": 190
},
{
"dpo_loss": 0.535432755947113,
"epoch": 0.28044103547459254,
"grad_norm": 14.462798502305684,
"learning_rate": 4.522542485937369e-06,
"logits": -1.5289549827575684,
"logps": -82.07774353027344,
"loss": 0.5768,
"objective": 0.535432755947113,
"ranking_simple": 0.6958333253860474,
"step": 195
},
{
"dpo_loss": 0.664901614189148,
"epoch": 0.28763183125599234,
"grad_norm": 14.055996226374166,
"learning_rate": 4.4849759966195885e-06,
"logits": -1.4697571992874146,
"logps": -84.17499542236328,
"loss": 0.6108,
"objective": 0.664901614189148,
"ranking_simple": 0.5291666388511658,
"step": 200
},
{
"epoch": 0.28763183125599234,
"eval_dpo_loss": 0.6794703006744385,
"eval_logits": -1.406295895576477,
"eval_logps": -85.59771728515625,
"eval_loss": 0.6792395710945129,
"eval_objective": 0.6794703006744385,
"eval_ranking_simple": 0.5467567443847656,
"eval_runtime": 672.3535,
"eval_samples_per_second": 16.493,
"eval_steps_per_second": 1.376,
"step": 200
},
{
"dpo_loss": 0.5508471131324768,
"epoch": 0.29482262703739215,
"grad_norm": 12.878307216053773,
"learning_rate": 4.446155753917559e-06,
"logits": -1.4630080461502075,
"logps": -82.61222839355469,
"loss": 0.5818,
"objective": 0.5508471131324768,
"ranking_simple": 0.6166666746139526,
"step": 205
},
{
"dpo_loss": 0.6260824203491211,
"epoch": 0.30201342281879195,
"grad_norm": 13.972188885138191,
"learning_rate": 4.40610627752862e-06,
"logits": -1.5031265020370483,
"logps": -82.899658203125,
"loss": 0.6104,
"objective": 0.6260824203491211,
"ranking_simple": 0.5916666388511658,
"step": 210
},
{
"dpo_loss": 0.5617867708206177,
"epoch": 0.30920421860019176,
"grad_norm": 12.206297979335497,
"learning_rate": 4.364852863560456e-06,
"logits": -1.5797665119171143,
"logps": -81.89352416992188,
"loss": 0.5669,
"objective": 0.5617867708206177,
"ranking_simple": 0.6083333492279053,
"step": 215
},
{
"dpo_loss": 0.4900062680244446,
"epoch": 0.31639501438159157,
"grad_norm": 13.75475301520002,
"learning_rate": 4.322421568553529e-06,
"logits": -1.571911096572876,
"logps": -85.30619812011719,
"loss": 0.5582,
"objective": 0.4900062680244446,
"ranking_simple": 0.6499999761581421,
"step": 220
},
{
"dpo_loss": 0.6220059990882874,
"epoch": 0.32358581016299137,
"grad_norm": 14.857535963350637,
"learning_rate": 4.278839193023214e-06,
"logits": -1.6418553590774536,
"logps": -86.45494842529297,
"loss": 0.5991,
"objective": 0.6220059990882874,
"ranking_simple": 0.5833333134651184,
"step": 225
},
{
"dpo_loss": 0.5970975160598755,
"epoch": 0.3307766059443912,
"grad_norm": 12.612318414450316,
"learning_rate": 4.234133264532012e-06,
"logits": -1.5469430685043335,
"logps": -83.91010284423828,
"loss": 0.5904,
"objective": 0.5970975160598755,
"ranking_simple": 0.574999988079071,
"step": 230
},
{
"dpo_loss": 0.5481938123703003,
"epoch": 0.337967401725791,
"grad_norm": 12.797277404018391,
"learning_rate": 4.188332020302561e-06,
"logits": -1.3681390285491943,
"logps": -83.8482894897461,
"loss": 0.5406,
"objective": 0.5481938123703003,
"ranking_simple": 0.6499999761581421,
"step": 235
},
{
"dpo_loss": 0.6017696261405945,
"epoch": 0.3451581975071908,
"grad_norm": 13.86217536920719,
"learning_rate": 4.141464389382392e-06,
"logits": -1.4286202192306519,
"logps": -85.05973052978516,
"loss": 0.5855,
"objective": 0.6017696261405945,
"ranking_simple": 0.6041666865348816,
"step": 240
},
{
"dpo_loss": 0.5819531679153442,
"epoch": 0.3523489932885906,
"grad_norm": 13.117346601405215,
"learning_rate": 4.093559974371725e-06,
"logits": -1.4333292245864868,
"logps": -85.2267837524414,
"loss": 0.5903,
"objective": 0.5819531679153442,
"ranking_simple": 0.5666666626930237,
"step": 245
},
{
"dpo_loss": 0.6078858971595764,
"epoch": 0.3595397890699904,
"grad_norm": 12.057808368664894,
"learning_rate": 4.044649032725836e-06,
"logits": -1.43559992313385,
"logps": -83.82780456542969,
"loss": 0.5856,
"objective": 0.6078858971595764,
"ranking_simple": 0.6041666865348816,
"step": 250
},
{
"epoch": 0.3595397890699904,
"eval_dpo_loss": 0.6614840626716614,
"eval_logits": -1.3857855796813965,
"eval_logps": -87.80554962158203,
"eval_loss": 0.6622549295425415,
"eval_objective": 0.6614840626716614,
"eval_ranking_simple": 0.545945942401886,
"eval_runtime": 667.2429,
"eval_samples_per_second": 16.619,
"eval_steps_per_second": 1.386,
"step": 250
},
{
"dpo_loss": 0.47653335332870483,
"epoch": 0.3667305848513902,
"grad_norm": 12.396454934600511,
"learning_rate": 3.9947624576437975e-06,
"logits": -1.3619633913040161,
"logps": -85.24054718017578,
"loss": 0.5457,
"objective": 0.47653335332870483,
"ranking_simple": 0.6499999761581421,
"step": 255
},
{
"dpo_loss": 0.4943782687187195,
"epoch": 0.37392138063279,
"grad_norm": 12.89782329658142,
"learning_rate": 3.943931758555669e-06,
"logits": -1.4246817827224731,
"logps": -85.02723693847656,
"loss": 0.5225,
"objective": 0.4943782687187195,
"ranking_simple": 0.6458333134651184,
"step": 260
},
{
"dpo_loss": 0.6019367575645447,
"epoch": 0.3811121764141898,
"grad_norm": 14.075516870638713,
"learning_rate": 3.89218904122047e-06,
"logits": -1.4607974290847778,
"logps": -84.44505310058594,
"loss": 0.5847,
"objective": 0.6019367575645447,
"ranking_simple": 0.6166666746139526,
"step": 265
},
{
"dpo_loss": 0.5739615559577942,
"epoch": 0.3883029721955896,
"grad_norm": 13.302239992029932,
"learning_rate": 3.839566987447492e-06,
"logits": -1.4933629035949707,
"logps": -85.0285873413086,
"loss": 0.586,
"objective": 0.5739615559577942,
"ranking_simple": 0.6416666507720947,
"step": 270
},
{
"dpo_loss": 0.6036986708641052,
"epoch": 0.39549376797698943,
"grad_norm": 12.980951698600105,
"learning_rate": 3.7860988344537664e-06,
"logits": -1.536937952041626,
"logps": -85.6275863647461,
"loss": 0.5653,
"objective": 0.6036986708641052,
"ranking_simple": 0.6708333492279053,
"step": 275
},
{
"dpo_loss": 0.5608472228050232,
"epoch": 0.40268456375838924,
"grad_norm": 12.026354566513733,
"learning_rate": 3.731818353870729e-06,
"logits": -1.42515230178833,
"logps": -83.92908477783203,
"loss": 0.5729,
"objective": 0.5608472228050232,
"ranking_simple": 0.6208333373069763,
"step": 280
},
{
"dpo_loss": 0.5649358034133911,
"epoch": 0.4098753595397891,
"grad_norm": 12.506947473488001,
"learning_rate": 3.6767598304133325e-06,
"logits": -1.386040210723877,
"logps": -83.64810180664062,
"loss": 0.5379,
"objective": 0.5649358034133911,
"ranking_simple": 0.6458333134651184,
"step": 285
},
{
"dpo_loss": 0.5191709399223328,
"epoch": 0.4170661553211889,
"grad_norm": 11.83713099591173,
"learning_rate": 3.6209580402250816e-06,
"logits": -1.464784026145935,
"logps": -82.18307495117188,
"loss": 0.5179,
"objective": 0.5191709399223328,
"ranking_simple": 0.5958333611488342,
"step": 290
},
{
"dpo_loss": 0.5659650564193726,
"epoch": 0.4242569511025887,
"grad_norm": 13.171214716747091,
"learning_rate": 3.564448228912682e-06,
"logits": -1.456804633140564,
"logps": -85.0264892578125,
"loss": 0.56,
"objective": 0.5659650564193726,
"ranking_simple": 0.637499988079071,
"step": 295
},
{
"dpo_loss": 0.5784734487533569,
"epoch": 0.4314477468839885,
"grad_norm": 14.743575255965931,
"learning_rate": 3.507266089284157e-06,
"logits": -1.4471509456634521,
"logps": -86.41505432128906,
"loss": 0.5701,
"objective": 0.5784734487533569,
"ranking_simple": 0.6541666388511658,
"step": 300
},
{
"epoch": 0.4314477468839885,
"eval_dpo_loss": 0.6661913990974426,
"eval_logits": -1.478379487991333,
"eval_logps": -88.78189086914062,
"eval_loss": 0.6633257269859314,
"eval_objective": 0.6661913990974426,
"eval_ranking_simple": 0.5448648929595947,
"eval_runtime": 671.0,
"eval_samples_per_second": 16.526,
"eval_steps_per_second": 1.379,
"step": 300
},
{
"dpo_loss": 0.6009036302566528,
"epoch": 0.4386385426653883,
"grad_norm": 13.785096580105144,
"learning_rate": 3.4494477388045035e-06,
"logits": -1.5999183654785156,
"logps": -88.15947723388672,
"loss": 0.5422,
"objective": 0.6009036302566528,
"ranking_simple": 0.625,
"step": 305
},
{
"dpo_loss": 0.4983634650707245,
"epoch": 0.4458293384467881,
"grad_norm": 12.441589802769128,
"learning_rate": 3.391029696783127e-06,
"logits": -1.5172799825668335,
"logps": -85.33808135986328,
"loss": 0.5392,
"objective": 0.4983634650707245,
"ranking_simple": 0.6708333492279053,
"step": 310
},
{
"dpo_loss": 0.48155930638313293,
"epoch": 0.45302013422818793,
"grad_norm": 13.232427609955911,
"learning_rate": 3.332048861307467e-06,
"logits": -1.5914006233215332,
"logps": -86.23084259033203,
"loss": 0.5352,
"objective": 0.48155930638313293,
"ranking_simple": 0.6666666865348816,
"step": 315
},
{
"dpo_loss": 0.5351694226264954,
"epoch": 0.46021093000958774,
"grad_norm": 12.016279548789747,
"learning_rate": 3.272542485937369e-06,
"logits": -1.5048084259033203,
"logps": -86.4006118774414,
"loss": 0.5458,
"objective": 0.5351694226264954,
"ranking_simple": 0.637499988079071,
"step": 320
},
{
"dpo_loss": 0.6042866110801697,
"epoch": 0.46740172579098754,
"grad_norm": 11.801563225543802,
"learning_rate": 3.2125481561749406e-06,
"logits": -1.5276148319244385,
"logps": -88.05933380126953,
"loss": 0.5391,
"objective": 0.6042866110801697,
"ranking_simple": 0.6166666746139526,
"step": 325
},
{
"dpo_loss": 0.5653389096260071,
"epoch": 0.47459252157238735,
"grad_norm": 13.17915800680032,
"learning_rate": 3.152103765724743e-06,
"logits": -1.6013100147247314,
"logps": -87.7344741821289,
"loss": 0.5488,
"objective": 0.5653389096260071,
"ranking_simple": 0.6583333611488342,
"step": 330
},
{
"dpo_loss": 0.5424811840057373,
"epoch": 0.48178331735378715,
"grad_norm": 11.944202243524137,
"learning_rate": 3.0912474925593124e-06,
"logits": -1.625429391860962,
"logps": -87.23538208007812,
"loss": 0.5222,
"objective": 0.5424811840057373,
"ranking_simple": 0.6041666865348816,
"step": 335
},
{
"dpo_loss": 0.5668550729751587,
"epoch": 0.48897411313518696,
"grad_norm": 13.39845062211121,
"learning_rate": 3.0300177748051375e-06,
"logits": -1.6717287302017212,
"logps": -87.84339904785156,
"loss": 0.5531,
"objective": 0.5668550729751587,
"ranking_simple": 0.6166666746139526,
"step": 340
},
{
"dpo_loss": 0.5499953627586365,
"epoch": 0.49616490891658677,
"grad_norm": 13.090607166084366,
"learning_rate": 2.9684532864643123e-06,
"logits": -1.593755841255188,
"logps": -85.349365234375,
"loss": 0.5647,
"objective": 0.5499953627586365,
"ranking_simple": 0.637499988079071,
"step": 345
},
{
"dpo_loss": 0.5135945081710815,
"epoch": 0.5033557046979866,
"grad_norm": 13.902934776591621,
"learning_rate": 2.9065929129872097e-06,
"logits": -1.5004137754440308,
"logps": -86.49576568603516,
"loss": 0.5296,
"objective": 0.5135945081710815,
"ranking_simple": 0.6583333611488342,
"step": 350
},
{
"epoch": 0.5033557046979866,
"eval_dpo_loss": 0.66639244556427,
"eval_logits": -1.5418288707733154,
"eval_logps": -89.54682159423828,
"eval_loss": 0.6642188429832458,
"eval_objective": 0.66639244556427,
"eval_ranking_simple": 0.5456756949424744,
"eval_runtime": 667.0903,
"eval_samples_per_second": 16.623,
"eval_steps_per_second": 1.387,
"step": 350
},
{
"dpo_loss": 0.5924364328384399,
"epoch": 0.5105465004793864,
"grad_norm": 15.350144877247981,
"learning_rate": 2.844475726711595e-06,
"logits": -1.5987753868103027,
"logps": -88.64606475830078,
"loss": 0.5333,
"objective": 0.5924364328384399,
"ranking_simple": 0.6416666507720947,
"step": 355
},
{
"dpo_loss": 0.5756219029426575,
"epoch": 0.5177372962607862,
"grad_norm": 13.527406313681999,
"learning_rate": 2.7821409621837042e-06,
"logits": -1.641775369644165,
"logps": -87.02297973632812,
"loss": 0.5326,
"objective": 0.5756219029426575,
"ranking_simple": 0.6166666746139526,
"step": 360
},
{
"dpo_loss": 0.5566696524620056,
"epoch": 0.524928092042186,
"grad_norm": 12.992979137134387,
"learning_rate": 2.7196279913768587e-06,
"logits": -1.6056040525436401,
"logps": -88.7591781616211,
"loss": 0.5132,
"objective": 0.5566696524620056,
"ranking_simple": 0.612500011920929,
"step": 365
},
{
"dpo_loss": 0.5310615301132202,
"epoch": 0.5321188878235859,
"grad_norm": 12.368172686048874,
"learning_rate": 2.6569762988232838e-06,
"logits": -1.6386042833328247,
"logps": -88.06958770751953,
"loss": 0.5415,
"objective": 0.5310615301132202,
"ranking_simple": 0.675000011920929,
"step": 370
},
{
"dpo_loss": 0.49322208762168884,
"epoch": 0.5393096836049857,
"grad_norm": 14.332678107736337,
"learning_rate": 2.594225456674837e-06,
"logits": -1.642881989479065,
"logps": -88.7568359375,
"loss": 0.5239,
"objective": 0.49322208762168884,
"ranking_simple": 0.6333333253860474,
"step": 375
},
{
"dpo_loss": 0.5207617878913879,
"epoch": 0.5465004793863855,
"grad_norm": 12.666539980855852,
"learning_rate": 2.531415099708382e-06,
"logits": -1.6033991575241089,
"logps": -87.1814956665039,
"loss": 0.5323,
"objective": 0.5207617878913879,
"ranking_simple": 0.6416666507720947,
"step": 380
},
{
"dpo_loss": 0.5247195363044739,
"epoch": 0.5536912751677853,
"grad_norm": 12.467690022940683,
"learning_rate": 2.4685849002916184e-06,
"logits": -1.5490994453430176,
"logps": -87.7108154296875,
"loss": 0.5402,
"objective": 0.5247195363044739,
"ranking_simple": 0.6625000238418579,
"step": 385
},
{
"dpo_loss": 0.5364466309547424,
"epoch": 0.5608820709491851,
"grad_norm": 13.247530100964932,
"learning_rate": 2.4057745433251637e-06,
"logits": -1.5940817594528198,
"logps": -83.6358871459961,
"loss": 0.5439,
"objective": 0.5364466309547424,
"ranking_simple": 0.5791666507720947,
"step": 390
},
{
"dpo_loss": 0.5119186639785767,
"epoch": 0.5680728667305849,
"grad_norm": 13.353391264425635,
"learning_rate": 2.3430237011767166e-06,
"logits": -1.462990641593933,
"logps": -88.32820892333984,
"loss": 0.5228,
"objective": 0.5119186639785767,
"ranking_simple": 0.625,
"step": 395
},
{
"dpo_loss": 0.5097938776016235,
"epoch": 0.5752636625119847,
"grad_norm": 12.303363355643434,
"learning_rate": 2.280372008623142e-06,
"logits": -1.3568633794784546,
"logps": -87.81967163085938,
"loss": 0.5055,
"objective": 0.5097938776016235,
"ranking_simple": 0.5958333611488342,
"step": 400
},
{
"epoch": 0.5752636625119847,
"eval_dpo_loss": 0.6593486666679382,
"eval_logits": -1.4127254486083984,
"eval_logps": -89.56134796142578,
"eval_loss": 0.6579382419586182,
"eval_objective": 0.6593486666679382,
"eval_ranking_simple": 0.5535135269165039,
"eval_runtime": 673.313,
"eval_samples_per_second": 16.469,
"eval_steps_per_second": 1.374,
"step": 400
},
{
"dpo_loss": 0.5508550405502319,
"epoch": 0.5824544582933845,
"grad_norm": 12.24633763991367,
"learning_rate": 2.2178590378162957e-06,
"logits": -1.4666920900344849,
"logps": -84.19215393066406,
"loss": 0.5126,
"objective": 0.5508550405502319,
"ranking_simple": 0.6166666746139526,
"step": 405
},
{
"dpo_loss": 0.45802783966064453,
"epoch": 0.5896452540747843,
"grad_norm": 12.200895787251042,
"learning_rate": 2.155524273288405e-06,
"logits": -1.4165197610855103,
"logps": -86.24173736572266,
"loss": 0.4984,
"objective": 0.45802783966064453,
"ranking_simple": 0.6625000238418579,
"step": 410
},
{
"dpo_loss": 0.5100553035736084,
"epoch": 0.5968360498561841,
"grad_norm": 11.346657198974036,
"learning_rate": 2.093407087012791e-06,
"logits": -1.4899728298187256,
"logps": -85.83265686035156,
"loss": 0.4831,
"objective": 0.5100553035736084,
"ranking_simple": 0.6166666746139526,
"step": 415
},
{
"dpo_loss": 0.5085461139678955,
"epoch": 0.6040268456375839,
"grad_norm": 11.635360179200692,
"learning_rate": 2.031546713535688e-06,
"logits": -1.4201585054397583,
"logps": -86.8657455444336,
"loss": 0.5288,
"objective": 0.5085461139678955,
"ranking_simple": 0.6625000238418579,
"step": 420
},
{
"dpo_loss": 0.47269171476364136,
"epoch": 0.6112176414189837,
"grad_norm": 12.428264913670654,
"learning_rate": 1.969982225194864e-06,
"logits": -1.5006626844406128,
"logps": -86.10279083251953,
"loss": 0.5131,
"objective": 0.47269171476364136,
"ranking_simple": 0.6666666865348816,
"step": 425
},
{
"dpo_loss": 0.5131818056106567,
"epoch": 0.6184084372003835,
"grad_norm": 13.248218878689803,
"learning_rate": 1.908752507440689e-06,
"logits": -1.387397289276123,
"logps": -87.1451644897461,
"loss": 0.4929,
"objective": 0.5131818056106567,
"ranking_simple": 0.6291666626930237,
"step": 430
},
{
"dpo_loss": 0.5004034042358398,
"epoch": 0.6255992329817833,
"grad_norm": 13.300116401195462,
"learning_rate": 1.8478962342752584e-06,
"logits": -1.5423215627670288,
"logps": -88.39946746826172,
"loss": 0.4938,
"objective": 0.5004034042358398,
"ranking_simple": 0.7166666388511658,
"step": 435
},
{
"dpo_loss": 0.5937944054603577,
"epoch": 0.6327900287631831,
"grad_norm": 11.725152089523224,
"learning_rate": 1.7874518438250598e-06,
"logits": -1.5435981750488281,
"logps": -88.17759704589844,
"loss": 0.5132,
"objective": 0.5937944054603577,
"ranking_simple": 0.5874999761581421,
"step": 440
},
{
"dpo_loss": 0.4983668625354767,
"epoch": 0.6399808245445829,
"grad_norm": 12.259686808708759,
"learning_rate": 1.7274575140626318e-06,
"logits": -1.5366530418395996,
"logps": -89.67758178710938,
"loss": 0.4849,
"objective": 0.4983668625354767,
"ranking_simple": 0.6333333253860474,
"step": 445
},
{
"dpo_loss": 0.5500684380531311,
"epoch": 0.6471716203259827,
"grad_norm": 14.756271253525957,
"learning_rate": 1.6679511386925337e-06,
"logits": -1.5554006099700928,
"logps": -87.62480163574219,
"loss": 0.5493,
"objective": 0.5500684380531311,
"ranking_simple": 0.6541666388511658,
"step": 450
},
{
"epoch": 0.6471716203259827,
"eval_dpo_loss": 0.6628099083900452,
"eval_logits": -1.5076098442077637,
"eval_logps": -90.84712219238281,
"eval_loss": 0.6599166393280029,
"eval_objective": 0.6628099083900452,
"eval_ranking_simple": 0.5556756854057312,
"eval_runtime": 666.1599,
"eval_samples_per_second": 16.646,
"eval_steps_per_second": 1.389,
"step": 450
},
{
"dpo_loss": 0.46406909823417664,
"epoch": 0.6543624161073825,
"grad_norm": 15.01981163706973,
"learning_rate": 1.6089703032168736e-06,
"logits": -1.6114773750305176,
"logps": -85.98390197753906,
"loss": 0.4709,
"objective": 0.46406909823417664,
"ranking_simple": 0.6416666507720947,
"step": 455
},
{
"dpo_loss": 0.5671834945678711,
"epoch": 0.6615532118887824,
"grad_norm": 15.90483411128291,
"learning_rate": 1.5505522611954977e-06,
"logits": -1.611572265625,
"logps": -86.73653411865234,
"loss": 0.5229,
"objective": 0.5671834945678711,
"ranking_simple": 0.6499999761581421,
"step": 460
},
{
"dpo_loss": 0.49705228209495544,
"epoch": 0.6687440076701822,
"grad_norm": 12.71861951977229,
"learning_rate": 1.4927339107158437e-06,
"logits": -1.4986417293548584,
"logps": -86.50923156738281,
"loss": 0.5062,
"objective": 0.49705228209495544,
"ranking_simple": 0.6708333492279053,
"step": 465
},
{
"dpo_loss": 0.49713465571403503,
"epoch": 0.675934803451582,
"grad_norm": 13.38314630878278,
"learning_rate": 1.4355517710873184e-06,
"logits": -1.575962781906128,
"logps": -87.91844940185547,
"loss": 0.4998,
"objective": 0.49713465571403503,
"ranking_simple": 0.6625000238418579,
"step": 470
},
{
"dpo_loss": 0.4781602621078491,
"epoch": 0.6831255992329818,
"grad_norm": 14.056433998263255,
"learning_rate": 1.3790419597749198e-06,
"logits": -1.4974421262741089,
"logps": -86.79618835449219,
"loss": 0.5164,
"objective": 0.4781602621078491,
"ranking_simple": 0.6166666746139526,
"step": 475
},
{
"dpo_loss": 0.56535404920578,
"epoch": 0.6903163950143816,
"grad_norm": 14.229131759003886,
"learning_rate": 1.3232401695866686e-06,
"logits": -1.5553234815597534,
"logps": -88.59098052978516,
"loss": 0.5259,
"objective": 0.56535404920578,
"ranking_simple": 0.6541666388511658,
"step": 480
},
{
"dpo_loss": 0.5027016997337341,
"epoch": 0.6975071907957814,
"grad_norm": 13.871006060897372,
"learning_rate": 1.2681816461292715e-06,
"logits": -1.423956274986267,
"logps": -89.58472442626953,
"loss": 0.5072,
"objective": 0.5027016997337341,
"ranking_simple": 0.6791666746139526,
"step": 485
},
{
"dpo_loss": 0.50449138879776,
"epoch": 0.7046979865771812,
"grad_norm": 12.098287020865879,
"learning_rate": 1.2139011655462338e-06,
"logits": -1.48275887966156,
"logps": -91.03202819824219,
"loss": 0.4816,
"objective": 0.50449138879776,
"ranking_simple": 0.637499988079071,
"step": 490
},
{
"dpo_loss": 0.5269492268562317,
"epoch": 0.711888782358581,
"grad_norm": 12.367787564366719,
"learning_rate": 1.160433012552508e-06,
"logits": -1.6003237962722778,
"logps": -88.4117202758789,
"loss": 0.5101,
"objective": 0.5269492268562317,
"ranking_simple": 0.6291666626930237,
"step": 495
},
{
"dpo_loss": 0.48800936341285706,
"epoch": 0.7190795781399808,
"grad_norm": 14.023641814656244,
"learning_rate": 1.1078109587795311e-06,
"logits": -1.4825706481933594,
"logps": -87.40599060058594,
"loss": 0.5056,
"objective": 0.48800936341285706,
"ranking_simple": 0.5874999761581421,
"step": 500
},
{
"epoch": 0.7190795781399808,
"eval_dpo_loss": 0.659065842628479,
"eval_logits": -1.4782841205596924,
"eval_logps": -91.5462417602539,
"eval_loss": 0.656062662601471,
"eval_objective": 0.659065842628479,
"eval_ranking_simple": 0.5540540814399719,
"eval_runtime": 671.8381,
"eval_samples_per_second": 16.505,
"eval_steps_per_second": 1.377,
"step": 500
},
{
"dpo_loss": 0.5191890001296997,
"epoch": 0.7262703739213806,
"grad_norm": 14.79908104404648,
"learning_rate": 1.0560682414443315e-06,
"logits": -1.7025607824325562,
"logps": -88.41036224365234,
"loss": 0.5285,
"objective": 0.5191890001296997,
"ranking_simple": 0.6291666626930237,
"step": 505
},
{
"dpo_loss": 0.5311907529830933,
"epoch": 0.7334611697027804,
"grad_norm": 13.890053175833895,
"learning_rate": 1.0052375423562038e-06,
"logits": -1.5510889291763306,
"logps": -86.90328979492188,
"loss": 0.5148,
"objective": 0.5311907529830933,
"ranking_simple": 0.6708333492279053,
"step": 510
},
{
"dpo_loss": 0.5146838426589966,
"epoch": 0.7406519654841802,
"grad_norm": 13.696722595764014,
"learning_rate": 9.553509672741646e-07,
"logits": -1.498627781867981,
"logps": -89.00706481933594,
"loss": 0.5023,
"objective": 0.5146838426589966,
"ranking_simple": 0.6666666865348816,
"step": 515
},
{
"dpo_loss": 0.4828968048095703,
"epoch": 0.74784276126558,
"grad_norm": 14.629592063481873,
"learning_rate": 9.064400256282757e-07,
"logits": -1.5899360179901123,
"logps": -85.54163360595703,
"loss": 0.4789,
"objective": 0.4828968048095703,
"ranking_simple": 0.6458333134651184,
"step": 520
},
{
"dpo_loss": 0.4821651577949524,
"epoch": 0.7550335570469798,
"grad_norm": 12.203692188404089,
"learning_rate": 8.585356106176093e-07,
"logits": -1.5359408855438232,
"logps": -85.53140258789062,
"loss": 0.4681,
"objective": 0.4821651577949524,
"ranking_simple": 0.6708333492279053,
"step": 525
},
{
"dpo_loss": 0.4843992292881012,
"epoch": 0.7622243528283796,
"grad_norm": 14.26789388947485,
"learning_rate": 8.116679796974389e-07,
"logits": -1.5956252813339233,
"logps": -88.78003692626953,
"loss": 0.4665,
"objective": 0.4843992292881012,
"ranking_simple": 0.6916666626930237,
"step": 530
},
{
"dpo_loss": 0.48365145921707153,
"epoch": 0.7694151486097794,
"grad_norm": 14.283269123514527,
"learning_rate": 7.65866735467988e-07,
"logits": -1.56830632686615,
"logps": -87.40410614013672,
"loss": 0.5123,
"objective": 0.48365145921707153,
"ranking_simple": 0.637499988079071,
"step": 535
},
{
"dpo_loss": 0.4533773958683014,
"epoch": 0.7766059443911792,
"grad_norm": 12.484364657489317,
"learning_rate": 7.211608069767867e-07,
"logits": -1.5612220764160156,
"logps": -87.40184783935547,
"loss": 0.4608,
"objective": 0.4533773958683014,
"ranking_simple": 0.6625000238418579,
"step": 540
},
{
"dpo_loss": 0.5006808042526245,
"epoch": 0.783796740172579,
"grad_norm": 16.213825472630578,
"learning_rate": 6.775784314464717e-07,
"logits": -1.6361111402511597,
"logps": -86.7210464477539,
"loss": 0.5143,
"objective": 0.5006808042526245,
"ranking_simple": 0.6625000238418579,
"step": 545
},
{
"dpo_loss": 0.4449290633201599,
"epoch": 0.7909875359539789,
"grad_norm": 14.026431238273847,
"learning_rate": 6.351471364395448e-07,
"logits": -1.5031849145889282,
"logps": -86.18959045410156,
"loss": 0.475,
"objective": 0.4449290633201599,
"ranking_simple": 0.6875,
"step": 550
},
{
"epoch": 0.7909875359539789,
"eval_dpo_loss": 0.6598567962646484,
"eval_logits": -1.5167720317840576,
"eval_logps": -90.85450744628906,
"eval_loss": 0.6573106646537781,
"eval_objective": 0.6598567962646484,
"eval_ranking_simple": 0.5581080913543701,
"eval_runtime": 667.865,
"eval_samples_per_second": 16.604,
"eval_steps_per_second": 1.385,
"step": 550
},
{
"dpo_loss": 0.4697189927101135,
"epoch": 0.7981783317353787,
"grad_norm": 13.213685669112976,
"learning_rate": 5.9389372247138e-07,
"logits": -1.5569509267807007,
"logps": -89.16913604736328,
"loss": 0.4903,
"objective": 0.4697189927101135,
"ranking_simple": 0.675000011920929,
"step": 555
},
{
"dpo_loss": 0.4498293101787567,
"epoch": 0.8053691275167785,
"grad_norm": 13.815731473547668,
"learning_rate": 5.538442460824417e-07,
"logits": -1.6490610837936401,
"logps": -89.35672760009766,
"loss": 0.4906,
"objective": 0.4498293101787567,
"ranking_simple": 0.6666666865348816,
"step": 560
},
{
"dpo_loss": 0.47427651286125183,
"epoch": 0.8125599232981783,
"grad_norm": 13.523693416447607,
"learning_rate": 5.150240033804116e-07,
"logits": -1.617875337600708,
"logps": -87.59880065917969,
"loss": 0.4859,
"objective": 0.47427651286125183,
"ranking_simple": 0.6499999761581421,
"step": 565
},
{
"dpo_loss": 0.49262678623199463,
"epoch": 0.8197507190795782,
"grad_norm": 13.02579782169373,
"learning_rate": 4.774575140626317e-07,
"logits": -1.571083426475525,
"logps": -86.48162078857422,
"loss": 0.46,
"objective": 0.49262678623199463,
"ranking_simple": 0.6458333134651184,
"step": 570
},
{
"dpo_loss": 0.5198132991790771,
"epoch": 0.826941514860978,
"grad_norm": 11.87980002943589,
"learning_rate": 4.411685059289314e-07,
"logits": -1.5978747606277466,
"logps": -88.71498107910156,
"loss": 0.4784,
"objective": 0.5198132991790771,
"ranking_simple": 0.6458333134651184,
"step": 575
},
{
"dpo_loss": 0.46336984634399414,
"epoch": 0.8341323106423778,
"grad_norm": 15.96370979044969,
"learning_rate": 4.061798998946459e-07,
"logits": -1.5552499294281006,
"logps": -85.58879852294922,
"loss": 0.4752,
"objective": 0.46336984634399414,
"ranking_simple": 0.6916666626930237,
"step": 580
},
{
"dpo_loss": 0.41780319809913635,
"epoch": 0.8413231064237776,
"grad_norm": 13.134671346499106,
"learning_rate": 3.725137955132707e-07,
"logits": -1.6836026906967163,
"logps": -88.24311065673828,
"loss": 0.4566,
"objective": 0.41780319809913635,
"ranking_simple": 0.6708333492279053,
"step": 585
},
{
"dpo_loss": 0.4858356714248657,
"epoch": 0.8485139022051774,
"grad_norm": 13.954686906693901,
"learning_rate": 3.4019145701791186e-07,
"logits": -1.7384170293807983,
"logps": -88.8193359375,
"loss": 0.4763,
"objective": 0.4858356714248657,
"ranking_simple": 0.6708333492279053,
"step": 590
},
{
"dpo_loss": 0.48335954546928406,
"epoch": 0.8557046979865772,
"grad_norm": 13.802803355539663,
"learning_rate": 3.092332998903416e-07,
"logits": -1.6276822090148926,
"logps": -89.26622772216797,
"loss": 0.4727,
"objective": 0.48335954546928406,
"ranking_simple": 0.6583333611488342,
"step": 595
},
{
"dpo_loss": 0.522754430770874,
"epoch": 0.862895493767977,
"grad_norm": 13.432541739844822,
"learning_rate": 2.796588779661388e-07,
"logits": -1.5472941398620605,
"logps": -87.67240142822266,
"loss": 0.4881,
"objective": 0.522754430770874,
"ranking_simple": 0.6666666865348816,
"step": 600
},
{
"epoch": 0.862895493767977,
"eval_dpo_loss": 0.6609825491905212,
"eval_logits": -1.575743556022644,
"eval_logps": -91.89826202392578,
"eval_loss": 0.6581856608390808,
"eval_objective": 0.6609825491905212,
"eval_ranking_simple": 0.5564864873886108,
"eval_runtime": 667.1404,
"eval_samples_per_second": 16.622,
"eval_steps_per_second": 1.387,
"step": 600
},
{
"dpo_loss": 0.4695630371570587,
"epoch": 0.8700862895493768,
"grad_norm": 14.156954962823502,
"learning_rate": 2.514868710840723e-07,
"logits": -1.5538970232009888,
"logps": -86.31729888916016,
"loss": 0.4836,
"objective": 0.4695630371570587,
"ranking_simple": 0.6416666507720947,
"step": 605
},
{
"dpo_loss": 0.49627622961997986,
"epoch": 0.8772770853307766,
"grad_norm": 13.015536935516824,
"learning_rate": 2.2473507328751086e-07,
"logits": -1.6039773225784302,
"logps": -90.30487823486328,
"loss": 0.4714,
"objective": 0.49627622961997986,
"ranking_simple": 0.675000011920929,
"step": 610
},
{
"dpo_loss": 0.5044869184494019,
"epoch": 0.8844678811121764,
"grad_norm": 14.874997035211502,
"learning_rate": 1.9942038158532407e-07,
"logits": -1.6170308589935303,
"logps": -87.51586151123047,
"loss": 0.4845,
"objective": 0.5044869184494019,
"ranking_simple": 0.7333333492279053,
"step": 615
},
{
"dpo_loss": 0.4692457616329193,
"epoch": 0.8916586768935763,
"grad_norm": 14.018445832991219,
"learning_rate": 1.7555878527937164e-07,
"logits": -1.6254560947418213,
"logps": -85.93553924560547,
"loss": 0.4605,
"objective": 0.4692457616329193,
"ranking_simple": 0.7083333134651184,
"step": 620
},
{
"dpo_loss": 0.5408317446708679,
"epoch": 0.8988494726749761,
"grad_norm": 13.907456738178524,
"learning_rate": 1.5316535586531483e-07,
"logits": -1.5320008993148804,
"logps": -87.24546813964844,
"loss": 0.508,
"objective": 0.5408317446708679,
"ranking_simple": 0.675000011920929,
"step": 625
},
{
"dpo_loss": 0.4765617549419403,
"epoch": 0.9060402684563759,
"grad_norm": 11.837344650453549,
"learning_rate": 1.3225423751313942e-07,
"logits": -1.5696455240249634,
"logps": -88.36416625976562,
"loss": 0.4901,
"objective": 0.4765617549419403,
"ranking_simple": 0.6583333611488342,
"step": 630
},
{
"dpo_loss": 0.5072826743125916,
"epoch": 0.9132310642377757,
"grad_norm": 14.491002179336329,
"learning_rate": 1.1283863813339263e-07,
"logits": -1.571786880493164,
"logps": -87.18602752685547,
"loss": 0.5107,
"objective": 0.5072826743125916,
"ranking_simple": 0.5958333611488342,
"step": 635
},
{
"dpo_loss": 0.4825662672519684,
"epoch": 0.9204218600191755,
"grad_norm": 14.33764106585909,
"learning_rate": 9.493082103478519e-08,
"logits": -1.6901806592941284,
"logps": -89.96266174316406,
"loss": 0.4713,
"objective": 0.4825662672519684,
"ranking_simple": 0.6208333373069763,
"step": 640
},
{
"dpo_loss": 0.4537600874900818,
"epoch": 0.9276126558005753,
"grad_norm": 14.477057200242452,
"learning_rate": 7.854209717842231e-08,
"logits": -1.5616987943649292,
"logps": -89.00830841064453,
"loss": 0.465,
"objective": 0.4537600874900818,
"ranking_simple": 0.637499988079071,
"step": 645
},
{
"dpo_loss": 0.45157289505004883,
"epoch": 0.9348034515819751,
"grad_norm": 11.922257325246159,
"learning_rate": 6.368281803355692e-08,
"logits": -1.5918368101119995,
"logps": -87.64961242675781,
"loss": 0.4496,
"objective": 0.45157289505004883,
"ranking_simple": 0.6958333253860474,
"step": 650
},
{
"epoch": 0.9348034515819751,
"eval_dpo_loss": 0.6607991456985474,
"eval_logits": -1.5710175037384033,
"eval_logps": -91.90254211425781,
"eval_loss": 0.6576409935951233,
"eval_objective": 0.6607991456985474,
"eval_ranking_simple": 0.5572972893714905,
"eval_runtime": 667.4864,
"eval_samples_per_second": 16.613,
"eval_steps_per_second": 1.386,
"step": 650
},
{
"dpo_loss": 0.47816571593284607,
"epoch": 0.9419942473633749,
"grad_norm": 16.22501773533146,
"learning_rate": 5.036236903938285e-08,
"logits": -1.697513461112976,
"logps": -89.019287109375,
"loss": 0.4829,
"objective": 0.47816571593284607,
"ranking_simple": 0.6458333134651184,
"step": 655
},
{
"dpo_loss": 0.5012370347976685,
"epoch": 0.9491850431447747,
"grad_norm": 14.637112054873407,
"learning_rate": 3.8589163676986674e-08,
"logits": -1.6037873029708862,
"logps": -88.02587890625,
"loss": 0.4592,
"objective": 0.5012370347976685,
"ranking_simple": 0.6708333492279053,
"step": 660
},
{
"dpo_loss": 0.39999979734420776,
"epoch": 0.9563758389261745,
"grad_norm": 12.62525649619692,
"learning_rate": 2.8370638155215125e-08,
"logits": -1.6766146421432495,
"logps": -89.1705093383789,
"loss": 0.4545,
"objective": 0.39999979734420776,
"ranking_simple": 0.6625000238418579,
"step": 665
},
{
"dpo_loss": 0.40588125586509705,
"epoch": 0.9635666347075743,
"grad_norm": 13.937122960983983,
"learning_rate": 1.9713246713805588e-08,
"logits": -1.6513657569885254,
"logps": -89.80951690673828,
"loss": 0.4608,
"objective": 0.40588125586509705,
"ranking_simple": 0.675000011920929,
"step": 670
},
{
"dpo_loss": 0.5094369053840637,
"epoch": 0.9707574304889741,
"grad_norm": 14.410112405435674,
"learning_rate": 1.2622457546749567e-08,
"logits": -1.6709257364273071,
"logps": -90.41423034667969,
"loss": 0.4891,
"objective": 0.5094369053840637,
"ranking_simple": 0.6291666626930237,
"step": 675
},
{
"dpo_loss": 0.5134339928627014,
"epoch": 0.9779482262703739,
"grad_norm": 14.035267614600281,
"learning_rate": 7.102749348465166e-09,
"logits": -1.580245852470398,
"logps": -87.9394760131836,
"loss": 0.5082,
"objective": 0.5134339928627014,
"ranking_simple": 0.637499988079071,
"step": 680
},
{
"dpo_loss": 0.5040514469146729,
"epoch": 0.9851390220517737,
"grad_norm": 14.829160167874026,
"learning_rate": 3.1576084849563315e-09,
"logits": -1.5595978498458862,
"logps": -88.81024932861328,
"loss": 0.5014,
"objective": 0.5040514469146729,
"ranking_simple": 0.6458333134651184,
"step": 685
},
{
"dpo_loss": 0.5098596215248108,
"epoch": 0.9923298178331735,
"grad_norm": 15.252382359007699,
"learning_rate": 7.895267917501503e-10,
"logits": -1.6206508874893188,
"logps": -88.91633605957031,
"loss": 0.4697,
"objective": 0.5098596215248108,
"ranking_simple": 0.6583333611488342,
"step": 690
},
{
"dpo_loss": 0.47479790449142456,
"epoch": 0.9995206136145733,
"grad_norm": 14.700651928280335,
"learning_rate": 0.0,
"logits": -1.5682662725448608,
"logps": -88.02741241455078,
"loss": 0.4579,
"objective": 0.47479790449142456,
"ranking_simple": 0.6708333492279053,
"step": 695
},
{
"epoch": 0.9995206136145733,
"step": 695,
"total_flos": 0.0,
"train_loss": 0.5500740305125285,
"train_runtime": 23565.8136,
"train_samples_per_second": 4.249,
"train_steps_per_second": 0.029
}
],
"logging_steps": 5,
"max_steps": 695,
"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": 4,
"trial_name": null,
"trial_params": null
}