{
  "best_metric": 14.127137184143066,
  "best_model_checkpoint": "./qwen2.5-0.5b/qwen2.5-0.5b-expo-L2EXPO-ES-1/checkpoint-400",
  "epoch": 1.8422295701464337,
  "eval_steps": 50,
  "global_step": 650,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "dpo_loss": 0.6931471824645996,
      "epoch": 0.002834199338686821,
      "grad_norm": 368.84791774460115,
      "learning_rate": 2.840909090909091e-08,
      "logits": -1.359458565711975,
      "logps": -84.69721221923828,
      "loss": 0.3913,
      "objective": 0.3618059456348419,
      "ranking_idealized": 0.5833333134651184,
      "ranking_idealized_expo": 0.5833333134651184,
      "ranking_simple": 0.5833333134651184,
      "regularize": 0.3618059456348419,
      "step": 1,
      "wo_beta": 14.830931663513184
    },
    {
      "dpo_loss": 0.6998967528343201,
      "epoch": 0.14170996693434104,
      "grad_norm": 396.7221362251878,
      "learning_rate": 1.4204545454545458e-06,
      "logits": -1.4534623622894287,
      "logps": -84.4651870727539,
      "loss": 0.6418,
      "objective": 0.6346251964569092,
      "ranking_idealized": 0.5289115905761719,
      "ranking_idealized_expo": 0.5221088528633118,
      "ranking_simple": 0.5246598720550537,
      "regularize": 0.6346251964569092,
      "step": 50,
      "wo_beta": 15.657191276550293
    },
    {
      "epoch": 0.14170996693434104,
      "eval_dpo_loss": 0.7479657530784607,
      "eval_logits": -1.4384208917617798,
      "eval_logps": -89.57877349853516,
      "eval_loss": 0.7368742823600769,
      "eval_objective": 0.7343389391899109,
      "eval_ranking_idealized": 0.5295031070709229,
      "eval_ranking_idealized_expo": 0.5212215185165405,
      "eval_ranking_simple": 0.5248447060585022,
      "eval_regularize": 0.7343389391899109,
      "eval_runtime": 308.869,
      "eval_samples_per_second": 18.746,
      "eval_steps_per_second": 1.564,
      "eval_wo_beta": 16.0413818359375,
      "step": 50
    },
    {
      "dpo_loss": 0.9691944122314453,
      "epoch": 0.2834199338686821,
      "grad_norm": 273.93971490638415,
      "learning_rate": 2.8409090909090916e-06,
      "logits": -1.3613545894622803,
      "logps": -82.8901596069336,
      "loss": 1.7208,
      "objective": 1.7239964008331299,
      "ranking_idealized": 0.5241666436195374,
      "ranking_idealized_expo": 0.5137500166893005,
      "ranking_simple": 0.5274999737739563,
      "regularize": 1.7239964008331299,
      "step": 100,
      "wo_beta": 15.285738945007324
    },
    {
      "epoch": 0.2834199338686821,
      "eval_dpo_loss": 1.0867407321929932,
      "eval_logits": -1.3167681694030762,
      "eval_logps": -87.80636596679688,
      "eval_loss": 1.7081643342971802,
      "eval_objective": 1.6949896812438965,
      "eval_ranking_idealized": 0.5295031070709229,
      "eval_ranking_idealized_expo": 0.5212215185165405,
      "eval_ranking_simple": 0.522774338722229,
      "eval_regularize": 1.6949896812438965,
      "eval_runtime": 307.8806,
      "eval_samples_per_second": 18.806,
      "eval_steps_per_second": 1.569,
      "eval_wo_beta": 15.514751434326172,
      "step": 100
    },
    {
      "dpo_loss": 1.4132683277130127,
      "epoch": 0.42512990080302315,
      "grad_norm": 224.02387198038068,
      "learning_rate": 4.2613636363636365e-06,
      "logits": -1.1957015991210938,
      "logps": -80.9156723022461,
      "loss": 2.841,
      "objective": 2.8095057010650635,
      "ranking_idealized": 0.5333333611488342,
      "ranking_idealized_expo": 0.527916669845581,
      "ranking_simple": 0.5325000286102295,
      "regularize": 2.8095057010650635,
      "step": 150,
      "wo_beta": 15.042305946350098
    },
    {
      "epoch": 0.42512990080302315,
      "eval_dpo_loss": 1.6352450847625732,
      "eval_logits": -1.1085715293884277,
      "eval_logps": -83.1790771484375,
      "eval_loss": 2.930232048034668,
      "eval_objective": 2.876790761947632,
      "eval_ranking_idealized": 0.5295031070709229,
      "eval_ranking_idealized_expo": 0.5212215185165405,
      "eval_ranking_simple": 0.5300207138061523,
      "eval_regularize": 2.876790761947632,
      "eval_runtime": 311.4753,
      "eval_samples_per_second": 18.589,
      "eval_steps_per_second": 1.551,
      "eval_wo_beta": 15.068045616149902,
      "step": 150
    },
    {
      "dpo_loss": 1.8845717906951904,
      "epoch": 0.5668398677373642,
      "grad_norm": 197.31465488920477,
      "learning_rate": 4.997168347957521e-06,
      "logits": -0.9064983129501343,
      "logps": -77.56580352783203,
      "loss": 3.5072,
      "objective": 3.5702998638153076,
      "ranking_idealized": 0.5204166769981384,
      "ranking_idealized_expo": 0.51541668176651,
      "ranking_simple": 0.5104166865348816,
      "regularize": 3.5702998638153076,
      "step": 200,
      "wo_beta": 15.2503023147583
    },
    {
      "epoch": 0.5668398677373642,
      "eval_dpo_loss": 2.312016725540161,
      "eval_logits": -0.8688302040100098,
      "eval_logps": -80.29598236083984,
      "eval_loss": 4.2316670417785645,
      "eval_objective": 4.220970153808594,
      "eval_ranking_idealized": 0.5295031070709229,
      "eval_ranking_idealized_expo": 0.5212215185165405,
      "eval_ranking_simple": 0.5155279636383057,
      "eval_regularize": 4.220970153808594,
      "eval_runtime": 307.8579,
      "eval_samples_per_second": 18.807,
      "eval_steps_per_second": 1.569,
      "eval_wo_beta": 14.531935691833496,
      "step": 200
    },
    {
      "dpo_loss": 1.959755301475525,
      "epoch": 0.7085498346717053,
      "grad_norm": 180.7425737123104,
      "learning_rate": 4.973122855144066e-06,
      "logits": -0.8095456957817078,
      "logps": -75.73948669433594,
      "loss": 3.7707,
      "objective": 3.773972272872925,
      "ranking_idealized": 0.5249999761581421,
      "ranking_idealized_expo": 0.5162500143051147,
      "ranking_simple": 0.518750011920929,
      "regularize": 3.773972272872925,
      "step": 250,
      "wo_beta": 15.669358253479004
    },
    {
      "epoch": 0.7085498346717053,
      "eval_dpo_loss": 2.2987730503082275,
      "eval_logits": -0.7639057636260986,
      "eval_logps": -80.5389175415039,
      "eval_loss": 4.364786148071289,
      "eval_objective": 4.362744331359863,
      "eval_ranking_idealized": 0.5295031070709229,
      "eval_ranking_idealized_expo": 0.5212215185165405,
      "eval_ranking_simple": 0.5212215185165405,
      "eval_regularize": 4.362744331359863,
      "eval_runtime": 308.0276,
      "eval_samples_per_second": 18.797,
      "eval_steps_per_second": 1.568,
      "eval_wo_beta": 14.566258430480957,
      "step": 250
    },
    {
      "dpo_loss": 1.8572473526000977,
      "epoch": 0.8502598016060463,
      "grad_norm": 169.7352822578473,
      "learning_rate": 4.924776641419513e-06,
      "logits": -0.485324501991272,
      "logps": -79.25447082519531,
      "loss": 3.5773,
      "objective": 3.6851494312286377,
      "ranking_idealized": 0.5062500238418579,
      "ranking_idealized_expo": 0.4950000047683716,
      "ranking_simple": 0.5112500190734863,
      "regularize": 3.6851494312286377,
      "step": 300,
      "wo_beta": 15.083699226379395
    },
    {
      "epoch": 0.8502598016060463,
      "eval_dpo_loss": 2.2955195903778076,
      "eval_logits": -0.5387536883354187,
      "eval_logps": -83.8565444946289,
      "eval_loss": 4.3904218673706055,
      "eval_objective": 4.397186279296875,
      "eval_ranking_idealized": 0.5295031070709229,
      "eval_ranking_idealized_expo": 0.5212215185165405,
      "eval_ranking_simple": 0.523809552192688,
      "eval_regularize": 4.397186279296875,
      "eval_runtime": 308.1076,
      "eval_samples_per_second": 18.792,
      "eval_steps_per_second": 1.568,
      "eval_wo_beta": 14.309849739074707,
      "step": 300
    },
    {
      "dpo_loss": 1.6745129823684692,
      "epoch": 0.9919697685403873,
      "grad_norm": 175.43829028737628,
      "learning_rate": 4.8526047530778175e-06,
      "logits": -0.49194690585136414,
      "logps": -78.66015625,
      "loss": 3.359,
      "objective": 3.3252944946289062,
      "ranking_idealized": 0.5354166626930237,
      "ranking_idealized_expo": 0.5254166722297668,
      "ranking_simple": 0.5391666889190674,
      "regularize": 3.3252944946289062,
      "step": 350,
      "wo_beta": 15.092531204223633
    },
    {
      "epoch": 0.9919697685403873,
      "eval_dpo_loss": 2.417555809020996,
      "eval_logits": -0.555489718914032,
      "eval_logps": -82.1212387084961,
      "eval_loss": 4.6867547035217285,
      "eval_objective": 4.629337787628174,
      "eval_ranking_idealized": 0.5295031070709229,
      "eval_ranking_idealized_expo": 0.5212215185165405,
      "eval_ranking_simple": 0.5263975262641907,
      "eval_regularize": 4.629337787628174,
      "eval_runtime": 307.7738,
      "eval_samples_per_second": 18.813,
      "eval_steps_per_second": 1.569,
      "eval_wo_beta": 14.317716598510742,
      "step": 350
    },
    {
      "dpo_loss": 1.6242923736572266,
      "epoch": 1.1336797354747283,
      "grad_norm": 167.92800627071566,
      "learning_rate": 4.757316345716554e-06,
      "logits": -0.42070272564888,
      "logps": -77.89451599121094,
      "loss": 3.0892,
      "objective": 3.102174997329712,
      "ranking_idealized": 0.5412499904632568,
      "ranking_idealized_expo": 0.5320833325386047,
      "ranking_simple": 0.5333333611488342,
      "regularize": 3.102174997329712,
      "step": 400,
      "wo_beta": 15.249672889709473
    },
    {
      "epoch": 1.1336797354747283,
      "eval_dpo_loss": 2.573197364807129,
      "eval_logits": -0.484580934047699,
      "eval_logps": -80.18506622314453,
      "eval_loss": 4.899093151092529,
      "eval_objective": 4.920805931091309,
      "eval_ranking_idealized": 0.5295031070709229,
      "eval_ranking_idealized_expo": 0.5212215185165405,
      "eval_ranking_simple": 0.523809552192688,
      "eval_regularize": 4.920805931091309,
      "eval_runtime": 307.8724,
      "eval_samples_per_second": 18.806,
      "eval_steps_per_second": 1.569,
      "eval_wo_beta": 14.127137184143066,
      "step": 400
    },
    {
      "dpo_loss": 1.5614336729049683,
      "epoch": 1.2753897024090695,
      "grad_norm": 168.51104996086647,
      "learning_rate": 4.639847716126855e-06,
      "logits": -0.46561330556869507,
      "logps": -76.85901641845703,
      "loss": 3.001,
      "objective": 3.0770418643951416,
      "ranking_idealized": 0.5245833396911621,
      "ranking_idealized_expo": 0.5191666483879089,
      "ranking_simple": 0.5337499976158142,
      "regularize": 3.0770418643951416,
      "step": 450,
      "wo_beta": 15.840205192565918
    },
    {
      "epoch": 1.2753897024090695,
      "eval_dpo_loss": 2.4965932369232178,
      "eval_logits": -0.5097361207008362,
      "eval_logps": -82.07725524902344,
      "eval_loss": 4.865055561065674,
      "eval_objective": 4.8037590980529785,
      "eval_ranking_idealized": 0.5295031070709229,
      "eval_ranking_idealized_expo": 0.5212215185165405,
      "eval_ranking_simple": 0.5232919454574585,
      "eval_regularize": 4.8037590980529785,
      "eval_runtime": 308.1386,
      "eval_samples_per_second": 18.79,
      "eval_steps_per_second": 1.567,
      "eval_wo_beta": 14.230860710144043,
      "step": 450
    },
    {
      "dpo_loss": 1.4661251306533813,
      "epoch": 1.4170996693434104,
      "grad_norm": 153.41700493191905,
      "learning_rate": 4.501353102310901e-06,
      "logits": -0.41387155652046204,
      "logps": -77.91703033447266,
      "loss": 2.8358,
      "objective": 2.893219232559204,
      "ranking_idealized": 0.5054166913032532,
      "ranking_idealized_expo": 0.4970833361148834,
      "ranking_simple": 0.5058333277702332,
      "regularize": 2.893219232559204,
      "step": 500,
      "wo_beta": 15.346451759338379
    },
    {
      "epoch": 1.4170996693434104,
      "eval_dpo_loss": 2.5684561729431152,
      "eval_logits": -0.4937358796596527,
      "eval_logps": -81.95916748046875,
      "eval_loss": 4.873396873474121,
      "eval_objective": 4.854368686676025,
      "eval_ranking_idealized": 0.5295031070709229,
      "eval_ranking_idealized_expo": 0.5212215185165405,
      "eval_ranking_simple": 0.5243270993232727,
      "eval_regularize": 4.854368686676025,
      "eval_runtime": 307.7716,
      "eval_samples_per_second": 18.813,
      "eval_steps_per_second": 1.569,
      "eval_wo_beta": 14.26622486114502,
      "step": 500
    },
    {
      "dpo_loss": 1.3315966129302979,
      "epoch": 1.5588096362777515,
      "grad_norm": 153.30075040252478,
      "learning_rate": 4.34319334202531e-06,
      "logits": -0.4832386374473572,
      "logps": -78.64344787597656,
      "loss": 2.6622,
      "objective": 2.6787972450256348,
      "ranking_idealized": 0.5199999809265137,
      "ranking_idealized_expo": 0.5108333230018616,
      "ranking_simple": 0.5249999761581421,
      "regularize": 2.6787972450256348,
      "step": 550,
      "wo_beta": 15.080702781677246
    },
    {
      "epoch": 1.5588096362777515,
      "eval_dpo_loss": 2.544066905975342,
      "eval_logits": -0.551251232624054,
      "eval_logps": -81.50196075439453,
      "eval_loss": 4.876008033752441,
      "eval_objective": 4.909796714782715,
      "eval_ranking_idealized": 0.5295031070709229,
      "eval_ranking_idealized_expo": 0.5212215185165405,
      "eval_ranking_simple": 0.5243270993232727,
      "eval_regularize": 4.909796714782715,
      "eval_runtime": 307.8096,
      "eval_samples_per_second": 18.81,
      "eval_steps_per_second": 1.569,
      "eval_wo_beta": 14.252228736877441,
      "step": 550
    },
    {
      "dpo_loss": 1.3174678087234497,
      "epoch": 1.7005196032120926,
      "grad_norm": 164.16680167227398,
      "learning_rate": 4.16692250129073e-06,
      "logits": -0.4034644663333893,
      "logps": -80.2961654663086,
      "loss": 2.5417,
      "objective": 2.5419461727142334,
      "ranking_idealized": 0.5220833420753479,
      "ranking_idealized_expo": 0.5149999856948853,
      "ranking_simple": 0.5220833420753479,
      "regularize": 2.5419461727142334,
      "step": 600,
      "wo_beta": 15.080598831176758
    },
    {
      "epoch": 1.7005196032120926,
      "eval_dpo_loss": 2.5863354206085205,
      "eval_logits": -0.5043439269065857,
      "eval_logps": -83.9180908203125,
      "eval_loss": 5.032442092895508,
      "eval_objective": 5.0250935554504395,
      "eval_ranking_idealized": 0.5295031070709229,
      "eval_ranking_idealized_expo": 0.5212215185165405,
      "eval_ranking_simple": 0.5258799195289612,
      "eval_regularize": 5.0250935554504395,
      "eval_runtime": 308.2064,
      "eval_samples_per_second": 18.786,
      "eval_steps_per_second": 1.567,
      "eval_wo_beta": 14.232461929321289,
      "step": 600
    },
    {
      "dpo_loss": 1.3203412294387817,
      "epoch": 1.8422295701464337,
      "grad_norm": 160.7017872607094,
      "learning_rate": 3.974272604254906e-06,
      "logits": -0.48519474267959595,
      "logps": -80.25173950195312,
      "loss": 2.435,
      "objective": 2.4747252464294434,
      "ranking_idealized": 0.5408333539962769,
      "ranking_idealized_expo": 0.527916669845581,
      "ranking_simple": 0.5333333611488342,
      "regularize": 2.4747252464294434,
      "step": 650,
      "wo_beta": 15.727615356445312
    },
    {
      "epoch": 1.8422295701464337,
      "eval_dpo_loss": 2.6193807125091553,
      "eval_logits": -0.49383923411369324,
      "eval_logps": -83.8819808959961,
      "eval_loss": 5.0286407470703125,
      "eval_objective": 5.001297950744629,
      "eval_ranking_idealized": 0.5295031070709229,
      "eval_ranking_idealized_expo": 0.5212215185165405,
      "eval_ranking_simple": 0.5196687579154968,
      "eval_regularize": 5.001297950744629,
      "eval_runtime": 307.7261,
      "eval_samples_per_second": 18.815,
      "eval_steps_per_second": 1.57,
      "eval_wo_beta": 14.250398635864258,
      "step": 650
    },
    {
      "epoch": 1.8422295701464337,
      "step": 650,
      "total_flos": 0.0,
      "train_loss": 2.767508169504312,
      "train_runtime": 17905.6458,
      "train_samples_per_second": 14.186,
      "train_steps_per_second": 0.098
    }
  ],
  "logging_steps": 50,
  "max_steps": 1760,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 5,
  "save_steps": 50,
  "stateful_callbacks": {
    "EarlyStoppingCallback": {
      "args": {
        "early_stopping_patience": 5,
        "early_stopping_threshold": 0.0
      },
      "attributes": {
        "early_stopping_patience_counter": 0
      }
    },
    "TrainerControl": {
      "args": {
        "should_epoch_stop": false,
        "should_evaluate": false,
        "should_log": false,
        "should_save": true,
        "should_training_stop": true
      },
      "attributes": {}
    }
  },
  "total_flos": 0.0,
  "train_batch_size": 4,
  "trial_name": null,
  "trial_params": null
}