{
  "best_metric": 2.7174136638641357,
  "best_model_checkpoint": "./models/kobart_16_5.6e-5_datav2_min30_lp5.0_temperature1.0/checkpoint-10000",
  "epoch": 5.0,
  "global_step": 13240,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.04,
      "learning_rate": 4.2296072507552865e-06,
      "loss": 6.9463,
      "step": 100
    },
    {
      "epoch": 0.08,
      "learning_rate": 8.459214501510573e-06,
      "loss": 3.1486,
      "step": 200
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.2688821752265861e-05,
      "loss": 2.8463,
      "step": 300
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.6918429003021146e-05,
      "loss": 2.687,
      "step": 400
    },
    {
      "epoch": 0.19,
      "learning_rate": 2.1148036253776434e-05,
      "loss": 2.5878,
      "step": 500
    },
    {
      "epoch": 0.23,
      "learning_rate": 2.5377643504531723e-05,
      "loss": 2.5439,
      "step": 600
    },
    {
      "epoch": 0.26,
      "learning_rate": 2.960725075528701e-05,
      "loss": 2.4939,
      "step": 700
    },
    {
      "epoch": 0.3,
      "learning_rate": 3.383685800604229e-05,
      "loss": 2.4266,
      "step": 800
    },
    {
      "epoch": 0.34,
      "learning_rate": 3.8066465256797584e-05,
      "loss": 2.4146,
      "step": 900
    },
    {
      "epoch": 0.38,
      "learning_rate": 4.229607250755287e-05,
      "loss": 2.376,
      "step": 1000
    },
    {
      "epoch": 0.42,
      "learning_rate": 4.6525679758308154e-05,
      "loss": 2.3883,
      "step": 1100
    },
    {
      "epoch": 0.45,
      "learning_rate": 5.0755287009063445e-05,
      "loss": 2.3586,
      "step": 1200
    },
    {
      "epoch": 0.49,
      "learning_rate": 5.498489425981873e-05,
      "loss": 2.3647,
      "step": 1300
    },
    {
      "epoch": 0.53,
      "learning_rate": 5.564283316549178e-05,
      "loss": 2.3665,
      "step": 1400
    },
    {
      "epoch": 0.57,
      "learning_rate": 5.517287680429674e-05,
      "loss": 2.3411,
      "step": 1500
    },
    {
      "epoch": 0.6,
      "learning_rate": 5.4702920443101714e-05,
      "loss": 2.3251,
      "step": 1600
    },
    {
      "epoch": 0.64,
      "learning_rate": 5.423296408190668e-05,
      "loss": 2.3253,
      "step": 1700
    },
    {
      "epoch": 0.68,
      "learning_rate": 5.376300772071164e-05,
      "loss": 2.2813,
      "step": 1800
    },
    {
      "epoch": 0.72,
      "learning_rate": 5.3293051359516615e-05,
      "loss": 2.2943,
      "step": 1900
    },
    {
      "epoch": 0.76,
      "learning_rate": 5.2823094998321586e-05,
      "loss": 2.2844,
      "step": 2000
    },
    {
      "epoch": 0.79,
      "learning_rate": 5.235313863712655e-05,
      "loss": 2.2827,
      "step": 2100
    },
    {
      "epoch": 0.83,
      "learning_rate": 5.188318227593152e-05,
      "loss": 2.273,
      "step": 2200
    },
    {
      "epoch": 0.87,
      "learning_rate": 5.1413225914736487e-05,
      "loss": 2.2586,
      "step": 2300
    },
    {
      "epoch": 0.91,
      "learning_rate": 5.094326955354146e-05,
      "loss": 2.2669,
      "step": 2400
    },
    {
      "epoch": 0.94,
      "learning_rate": 5.047331319234642e-05,
      "loss": 2.1997,
      "step": 2500
    },
    {
      "epoch": 0.98,
      "learning_rate": 5.000335683115139e-05,
      "loss": 2.1917,
      "step": 2600
    },
    {
      "epoch": 1.02,
      "learning_rate": 4.9533400469956365e-05,
      "loss": 2.1308,
      "step": 2700
    },
    {
      "epoch": 1.06,
      "learning_rate": 4.906344410876133e-05,
      "loss": 1.9878,
      "step": 2800
    },
    {
      "epoch": 1.1,
      "learning_rate": 4.8593487747566294e-05,
      "loss": 2.01,
      "step": 2900
    },
    {
      "epoch": 1.13,
      "learning_rate": 4.8123531386371266e-05,
      "loss": 2.0093,
      "step": 3000
    },
    {
      "epoch": 1.17,
      "learning_rate": 4.765357502517623e-05,
      "loss": 1.9826,
      "step": 3100
    },
    {
      "epoch": 1.21,
      "learning_rate": 4.71836186639812e-05,
      "loss": 2.0209,
      "step": 3200
    },
    {
      "epoch": 1.25,
      "learning_rate": 4.6713662302786166e-05,
      "loss": 1.9782,
      "step": 3300
    },
    {
      "epoch": 1.28,
      "learning_rate": 4.624370594159114e-05,
      "loss": 2.0308,
      "step": 3400
    },
    {
      "epoch": 1.32,
      "learning_rate": 4.577374958039611e-05,
      "loss": 1.9901,
      "step": 3500
    },
    {
      "epoch": 1.36,
      "learning_rate": 4.5303793219201074e-05,
      "loss": 1.9989,
      "step": 3600
    },
    {
      "epoch": 1.4,
      "learning_rate": 4.483383685800604e-05,
      "loss": 2.0022,
      "step": 3700
    },
    {
      "epoch": 1.44,
      "learning_rate": 4.436388049681101e-05,
      "loss": 2.0036,
      "step": 3800
    },
    {
      "epoch": 1.47,
      "learning_rate": 4.3893924135615974e-05,
      "loss": 2.0289,
      "step": 3900
    },
    {
      "epoch": 1.51,
      "learning_rate": 4.3423967774420946e-05,
      "loss": 2.0227,
      "step": 4000
    },
    {
      "epoch": 1.55,
      "learning_rate": 4.295401141322592e-05,
      "loss": 1.9819,
      "step": 4100
    },
    {
      "epoch": 1.59,
      "learning_rate": 4.248405505203088e-05,
      "loss": 2.0031,
      "step": 4200
    },
    {
      "epoch": 1.62,
      "learning_rate": 4.201409869083585e-05,
      "loss": 1.9959,
      "step": 4300
    },
    {
      "epoch": 1.66,
      "learning_rate": 4.154414232964082e-05,
      "loss": 1.9967,
      "step": 4400
    },
    {
      "epoch": 1.7,
      "learning_rate": 4.107418596844578e-05,
      "loss": 1.9797,
      "step": 4500
    },
    {
      "epoch": 1.74,
      "learning_rate": 4.0604229607250753e-05,
      "loss": 1.9502,
      "step": 4600
    },
    {
      "epoch": 1.77,
      "learning_rate": 4.0134273246055725e-05,
      "loss": 2.0092,
      "step": 4700
    },
    {
      "epoch": 1.81,
      "learning_rate": 3.9664316884860696e-05,
      "loss": 1.9867,
      "step": 4800
    },
    {
      "epoch": 1.85,
      "learning_rate": 3.919436052366566e-05,
      "loss": 1.9437,
      "step": 4900
    },
    {
      "epoch": 1.89,
      "learning_rate": 3.8724404162470625e-05,
      "loss": 1.9617,
      "step": 5000
    },
    {
      "epoch": 1.89,
      "eval_bleu1": 29.2237,
      "eval_bleu2": 16.8919,
      "eval_bleu3": 9.7826,
      "eval_bleu4": 5.4461,
      "eval_gen_len": 48.0676,
      "eval_loss": 2.6145970821380615,
      "eval_rouge1": 35.2828,
      "eval_rouge2": 12.4993,
      "eval_rougeL": 22.9894,
      "eval_runtime": 128.6929,
      "eval_samples_per_second": 3.334,
      "eval_steps_per_second": 0.031,
      "step": 5000
    },
    {
      "epoch": 1.93,
      "learning_rate": 3.82544478012756e-05,
      "loss": 1.9911,
      "step": 5100
    },
    {
      "epoch": 1.96,
      "learning_rate": 3.778449144008056e-05,
      "loss": 1.9941,
      "step": 5200
    },
    {
      "epoch": 2.0,
      "learning_rate": 3.7314535078885526e-05,
      "loss": 1.9854,
      "step": 5300
    },
    {
      "epoch": 2.04,
      "learning_rate": 3.6844578717690504e-05,
      "loss": 1.7062,
      "step": 5400
    },
    {
      "epoch": 2.08,
      "learning_rate": 3.637462235649547e-05,
      "loss": 1.728,
      "step": 5500
    },
    {
      "epoch": 2.11,
      "learning_rate": 3.590466599530044e-05,
      "loss": 1.7103,
      "step": 5600
    },
    {
      "epoch": 2.15,
      "learning_rate": 3.5434709634105405e-05,
      "loss": 1.7175,
      "step": 5700
    },
    {
      "epoch": 2.19,
      "learning_rate": 3.496475327291037e-05,
      "loss": 1.7124,
      "step": 5800
    },
    {
      "epoch": 2.23,
      "learning_rate": 3.449479691171534e-05,
      "loss": 1.7402,
      "step": 5900
    },
    {
      "epoch": 2.27,
      "learning_rate": 3.4024840550520305e-05,
      "loss": 1.7074,
      "step": 6000
    },
    {
      "epoch": 2.3,
      "learning_rate": 3.3554884189325276e-05,
      "loss": 1.7131,
      "step": 6100
    },
    {
      "epoch": 2.34,
      "learning_rate": 3.308492782813025e-05,
      "loss": 1.7293,
      "step": 6200
    },
    {
      "epoch": 2.38,
      "learning_rate": 3.261497146693521e-05,
      "loss": 1.713,
      "step": 6300
    },
    {
      "epoch": 2.42,
      "learning_rate": 3.2145015105740184e-05,
      "loss": 1.7057,
      "step": 6400
    },
    {
      "epoch": 2.45,
      "learning_rate": 3.167505874454515e-05,
      "loss": 1.7163,
      "step": 6500
    },
    {
      "epoch": 2.49,
      "learning_rate": 3.120510238335011e-05,
      "loss": 1.7463,
      "step": 6600
    },
    {
      "epoch": 2.53,
      "learning_rate": 3.0735146022155084e-05,
      "loss": 1.7164,
      "step": 6700
    },
    {
      "epoch": 2.57,
      "learning_rate": 3.0265189660960056e-05,
      "loss": 1.7407,
      "step": 6800
    },
    {
      "epoch": 2.61,
      "learning_rate": 2.9795233299765024e-05,
      "loss": 1.7219,
      "step": 6900
    },
    {
      "epoch": 2.64,
      "learning_rate": 2.9325276938569988e-05,
      "loss": 1.7553,
      "step": 7000
    },
    {
      "epoch": 2.68,
      "learning_rate": 2.8855320577374956e-05,
      "loss": 1.7446,
      "step": 7100
    },
    {
      "epoch": 2.72,
      "learning_rate": 2.8385364216179924e-05,
      "loss": 1.7557,
      "step": 7200
    },
    {
      "epoch": 2.76,
      "learning_rate": 2.7915407854984896e-05,
      "loss": 1.732,
      "step": 7300
    },
    {
      "epoch": 2.79,
      "learning_rate": 2.744545149378986e-05,
      "loss": 1.7261,
      "step": 7400
    },
    {
      "epoch": 2.83,
      "learning_rate": 2.6975495132594828e-05,
      "loss": 1.75,
      "step": 7500
    },
    {
      "epoch": 2.87,
      "learning_rate": 2.65055387713998e-05,
      "loss": 1.7101,
      "step": 7600
    },
    {
      "epoch": 2.91,
      "learning_rate": 2.6035582410204767e-05,
      "loss": 1.7265,
      "step": 7700
    },
    {
      "epoch": 2.95,
      "learning_rate": 2.5565626049009732e-05,
      "loss": 1.7091,
      "step": 7800
    },
    {
      "epoch": 2.98,
      "learning_rate": 2.5095669687814703e-05,
      "loss": 1.7309,
      "step": 7900
    },
    {
      "epoch": 3.02,
      "learning_rate": 2.462571332661967e-05,
      "loss": 1.5762,
      "step": 8000
    },
    {
      "epoch": 3.06,
      "learning_rate": 2.415575696542464e-05,
      "loss": 1.5125,
      "step": 8100
    },
    {
      "epoch": 3.1,
      "learning_rate": 2.3685800604229604e-05,
      "loss": 1.5083,
      "step": 8200
    },
    {
      "epoch": 3.13,
      "learning_rate": 2.3215844243034575e-05,
      "loss": 1.4978,
      "step": 8300
    },
    {
      "epoch": 3.17,
      "learning_rate": 2.2745887881839543e-05,
      "loss": 1.4629,
      "step": 8400
    },
    {
      "epoch": 3.21,
      "learning_rate": 2.227593152064451e-05,
      "loss": 1.5143,
      "step": 8500
    },
    {
      "epoch": 3.25,
      "learning_rate": 2.180597515944948e-05,
      "loss": 1.4975,
      "step": 8600
    },
    {
      "epoch": 3.29,
      "learning_rate": 2.1336018798254447e-05,
      "loss": 1.5022,
      "step": 8700
    },
    {
      "epoch": 3.32,
      "learning_rate": 2.0866062437059415e-05,
      "loss": 1.4942,
      "step": 8800
    },
    {
      "epoch": 3.36,
      "learning_rate": 2.0396106075864386e-05,
      "loss": 1.5088,
      "step": 8900
    },
    {
      "epoch": 3.4,
      "learning_rate": 1.992614971466935e-05,
      "loss": 1.5331,
      "step": 9000
    },
    {
      "epoch": 3.44,
      "learning_rate": 1.945619335347432e-05,
      "loss": 1.5323,
      "step": 9100
    },
    {
      "epoch": 3.47,
      "learning_rate": 1.8986236992279287e-05,
      "loss": 1.5066,
      "step": 9200
    },
    {
      "epoch": 3.51,
      "learning_rate": 1.851628063108426e-05,
      "loss": 1.523,
      "step": 9300
    },
    {
      "epoch": 3.55,
      "learning_rate": 1.8046324269889223e-05,
      "loss": 1.5198,
      "step": 9400
    },
    {
      "epoch": 3.59,
      "learning_rate": 1.757636790869419e-05,
      "loss": 1.541,
      "step": 9500
    },
    {
      "epoch": 3.63,
      "learning_rate": 1.7106411547499162e-05,
      "loss": 1.5227,
      "step": 9600
    },
    {
      "epoch": 3.66,
      "learning_rate": 1.663645518630413e-05,
      "loss": 1.5305,
      "step": 9700
    },
    {
      "epoch": 3.7,
      "learning_rate": 1.6166498825109095e-05,
      "loss": 1.5111,
      "step": 9800
    },
    {
      "epoch": 3.74,
      "learning_rate": 1.5696542463914063e-05,
      "loss": 1.4977,
      "step": 9900
    },
    {
      "epoch": 3.78,
      "learning_rate": 1.5226586102719034e-05,
      "loss": 1.5272,
      "step": 10000
    },
    {
      "epoch": 3.78,
      "eval_bleu1": 29.9954,
      "eval_bleu2": 17.513,
      "eval_bleu3": 10.317,
      "eval_bleu4": 5.8532,
      "eval_gen_len": 49.3147,
      "eval_loss": 2.7174136638641357,
      "eval_rouge1": 35.7621,
      "eval_rouge2": 12.8914,
      "eval_rougeL": 23.6695,
      "eval_runtime": 121.1732,
      "eval_samples_per_second": 3.54,
      "eval_steps_per_second": 0.033,
      "step": 10000
    },
    {
      "epoch": 3.81,
      "learning_rate": 1.4756629741524e-05,
      "loss": 1.5158,
      "step": 10100
    },
    {
      "epoch": 3.85,
      "learning_rate": 1.4286673380328968e-05,
      "loss": 1.5316,
      "step": 10200
    },
    {
      "epoch": 3.89,
      "learning_rate": 1.3816717019133936e-05,
      "loss": 1.5273,
      "step": 10300
    },
    {
      "epoch": 3.93,
      "learning_rate": 1.3346760657938906e-05,
      "loss": 1.5232,
      "step": 10400
    },
    {
      "epoch": 3.97,
      "learning_rate": 1.2876804296743874e-05,
      "loss": 1.5244,
      "step": 10500
    },
    {
      "epoch": 4.0,
      "learning_rate": 1.2406847935548842e-05,
      "loss": 1.5126,
      "step": 10600
    },
    {
      "epoch": 4.04,
      "learning_rate": 1.193689157435381e-05,
      "loss": 1.3597,
      "step": 10700
    },
    {
      "epoch": 4.08,
      "learning_rate": 1.1466935213158778e-05,
      "loss": 1.3505,
      "step": 10800
    },
    {
      "epoch": 4.12,
      "learning_rate": 1.0996978851963746e-05,
      "loss": 1.366,
      "step": 10900
    },
    {
      "epoch": 4.15,
      "learning_rate": 1.0527022490768716e-05,
      "loss": 1.3379,
      "step": 11000
    },
    {
      "epoch": 4.19,
      "learning_rate": 1.0057066129573682e-05,
      "loss": 1.3779,
      "step": 11100
    },
    {
      "epoch": 4.23,
      "learning_rate": 9.587109768378652e-06,
      "loss": 1.3452,
      "step": 11200
    },
    {
      "epoch": 4.27,
      "learning_rate": 9.117153407183618e-06,
      "loss": 1.3747,
      "step": 11300
    },
    {
      "epoch": 4.31,
      "learning_rate": 8.647197045988588e-06,
      "loss": 1.3467,
      "step": 11400
    },
    {
      "epoch": 4.34,
      "learning_rate": 8.177240684793554e-06,
      "loss": 1.3413,
      "step": 11500
    },
    {
      "epoch": 4.38,
      "learning_rate": 7.707284323598524e-06,
      "loss": 1.3677,
      "step": 11600
    },
    {
      "epoch": 4.42,
      "learning_rate": 7.2373279624034915e-06,
      "loss": 1.3461,
      "step": 11700
    },
    {
      "epoch": 4.46,
      "learning_rate": 6.7673716012084595e-06,
      "loss": 1.3679,
      "step": 11800
    },
    {
      "epoch": 4.49,
      "learning_rate": 6.2974152400134274e-06,
      "loss": 1.3608,
      "step": 11900
    },
    {
      "epoch": 4.53,
      "learning_rate": 5.827458878818395e-06,
      "loss": 1.3689,
      "step": 12000
    },
    {
      "epoch": 4.57,
      "learning_rate": 5.357502517623363e-06,
      "loss": 1.3679,
      "step": 12100
    },
    {
      "epoch": 4.61,
      "learning_rate": 4.887546156428331e-06,
      "loss": 1.3607,
      "step": 12200
    },
    {
      "epoch": 4.65,
      "learning_rate": 4.417589795233299e-06,
      "loss": 1.3842,
      "step": 12300
    },
    {
      "epoch": 4.68,
      "learning_rate": 3.947633434038268e-06,
      "loss": 1.3667,
      "step": 12400
    },
    {
      "epoch": 4.72,
      "learning_rate": 3.477677072843236e-06,
      "loss": 1.3444,
      "step": 12500
    },
    {
      "epoch": 4.76,
      "learning_rate": 3.007720711648204e-06,
      "loss": 1.3517,
      "step": 12600
    },
    {
      "epoch": 4.8,
      "learning_rate": 2.537764350453172e-06,
      "loss": 1.3818,
      "step": 12700
    },
    {
      "epoch": 4.83,
      "learning_rate": 2.0678079892581405e-06,
      "loss": 1.3711,
      "step": 12800
    },
    {
      "epoch": 4.87,
      "learning_rate": 1.5978516280631082e-06,
      "loss": 1.357,
      "step": 12900
    },
    {
      "epoch": 4.91,
      "learning_rate": 1.1278952668680764e-06,
      "loss": 1.3592,
      "step": 13000
    },
    {
      "epoch": 4.95,
      "learning_rate": 6.579389056730447e-07,
      "loss": 1.3488,
      "step": 13100
    },
    {
      "epoch": 4.98,
      "learning_rate": 1.8798254447801276e-07,
      "loss": 1.3811,
      "step": 13200
    },
    {
      "epoch": 5.0,
      "step": 13240,
      "total_flos": 3.139680217669632e+16,
      "train_loss": 1.8353108276413286,
      "train_runtime": 3683.8727,
      "train_samples_per_second": 57.503,
      "train_steps_per_second": 3.594
    }
  ],
  "max_steps": 13240,
  "num_train_epochs": 5,
  "total_flos": 3.139680217669632e+16,
  "trial_name": null,
  "trial_params": null
}