{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 5.0,
  "global_step": 37560,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.07,
      "learning_rate": 5e-05,
      "loss": 2.2617,
      "step": 500
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.932541824069077e-05,
      "loss": 1.6381,
      "step": 1000
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.865083648138155e-05,
      "loss": 1.4975,
      "step": 1500
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.7976254722072315e-05,
      "loss": 1.4772,
      "step": 2000
    },
    {
      "epoch": 0.33,
      "learning_rate": 4.730167296276309e-05,
      "loss": 1.4015,
      "step": 2500
    },
    {
      "epoch": 0.4,
      "learning_rate": 4.662709120345386e-05,
      "loss": 1.3626,
      "step": 3000
    },
    {
      "epoch": 0.47,
      "learning_rate": 4.595250944414463e-05,
      "loss": 1.3631,
      "step": 3500
    },
    {
      "epoch": 0.53,
      "learning_rate": 4.5277927684835404e-05,
      "loss": 1.3027,
      "step": 4000
    },
    {
      "epoch": 0.6,
      "learning_rate": 4.460334592552617e-05,
      "loss": 1.2998,
      "step": 4500
    },
    {
      "epoch": 0.67,
      "learning_rate": 4.392876416621695e-05,
      "loss": 1.2911,
      "step": 5000
    },
    {
      "epoch": 0.73,
      "learning_rate": 4.325418240690772e-05,
      "loss": 1.2764,
      "step": 5500
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.2579600647598486e-05,
      "loss": 1.2656,
      "step": 6000
    },
    {
      "epoch": 0.87,
      "learning_rate": 4.190501888828926e-05,
      "loss": 1.2599,
      "step": 6500
    },
    {
      "epoch": 0.93,
      "learning_rate": 4.123043712898003e-05,
      "loss": 1.2488,
      "step": 7000
    },
    {
      "epoch": 1.0,
      "learning_rate": 4.055585536967081e-05,
      "loss": 1.2533,
      "step": 7500
    },
    {
      "epoch": 1.0,
      "eval_bleu1_score": 0.9978,
      "eval_bleu2_score": 0.9978,
      "eval_bleu3_score": 0.9978,
      "eval_bleu4_score": 0.9978,
      "eval_loss": 1.2135246992111206,
      "eval_rougeL": 0.9976,
      "eval_runtime": 348.2564,
      "eval_samples_per_second": 4.795,
      "eval_steps_per_second": 2.398,
      "step": 7512
    },
    {
      "epoch": 1.06,
      "learning_rate": 3.988127361036158e-05,
      "loss": 1.0527,
      "step": 8000
    },
    {
      "epoch": 1.13,
      "learning_rate": 3.920669185105235e-05,
      "loss": 1.0452,
      "step": 8500
    },
    {
      "epoch": 1.2,
      "learning_rate": 3.8532110091743125e-05,
      "loss": 1.0486,
      "step": 9000
    },
    {
      "epoch": 1.26,
      "learning_rate": 3.7857528332433894e-05,
      "loss": 1.0376,
      "step": 9500
    },
    {
      "epoch": 1.33,
      "learning_rate": 3.718294657312467e-05,
      "loss": 1.05,
      "step": 10000
    },
    {
      "epoch": 1.4,
      "learning_rate": 3.650836481381544e-05,
      "loss": 1.0601,
      "step": 10500
    },
    {
      "epoch": 1.46,
      "learning_rate": 3.583378305450621e-05,
      "loss": 1.0451,
      "step": 11000
    },
    {
      "epoch": 1.53,
      "learning_rate": 3.515920129519698e-05,
      "loss": 1.0377,
      "step": 11500
    },
    {
      "epoch": 1.6,
      "learning_rate": 3.448461953588775e-05,
      "loss": 1.0407,
      "step": 12000
    },
    {
      "epoch": 1.66,
      "learning_rate": 3.3810037776578527e-05,
      "loss": 1.0325,
      "step": 12500
    },
    {
      "epoch": 1.73,
      "learning_rate": 3.3135456017269295e-05,
      "loss": 1.0092,
      "step": 13000
    },
    {
      "epoch": 1.8,
      "learning_rate": 3.2460874257960064e-05,
      "loss": 1.049,
      "step": 13500
    },
    {
      "epoch": 1.86,
      "learning_rate": 3.178629249865084e-05,
      "loss": 1.0264,
      "step": 14000
    },
    {
      "epoch": 1.93,
      "learning_rate": 3.111171073934161e-05,
      "loss": 1.0011,
      "step": 14500
    },
    {
      "epoch": 2.0,
      "learning_rate": 3.043712898003238e-05,
      "loss": 1.0329,
      "step": 15000
    },
    {
      "epoch": 2.0,
      "eval_bleu1_score": 0.9978,
      "eval_bleu2_score": 0.9978,
      "eval_bleu3_score": 0.9978,
      "eval_bleu4_score": 0.9978,
      "eval_loss": 1.1185928583145142,
      "eval_rougeL": 0.9976,
      "eval_runtime": 347.235,
      "eval_samples_per_second": 4.809,
      "eval_steps_per_second": 2.405,
      "step": 15024
    },
    {
      "epoch": 2.06,
      "learning_rate": 2.9762547220723153e-05,
      "loss": 0.8112,
      "step": 15500
    },
    {
      "epoch": 2.13,
      "learning_rate": 2.9087965461413925e-05,
      "loss": 0.8095,
      "step": 16000
    },
    {
      "epoch": 2.2,
      "learning_rate": 2.8413383702104697e-05,
      "loss": 0.8018,
      "step": 16500
    },
    {
      "epoch": 2.26,
      "learning_rate": 2.7738801942795465e-05,
      "loss": 0.7999,
      "step": 17000
    },
    {
      "epoch": 2.33,
      "learning_rate": 2.7064220183486238e-05,
      "loss": 0.8078,
      "step": 17500
    },
    {
      "epoch": 2.4,
      "learning_rate": 2.638963842417701e-05,
      "loss": 0.8053,
      "step": 18000
    },
    {
      "epoch": 2.46,
      "learning_rate": 2.5715056664867782e-05,
      "loss": 0.8008,
      "step": 18500
    },
    {
      "epoch": 2.53,
      "learning_rate": 2.5040474905558554e-05,
      "loss": 0.8187,
      "step": 19000
    },
    {
      "epoch": 2.6,
      "learning_rate": 2.4365893146249326e-05,
      "loss": 0.797,
      "step": 19500
    },
    {
      "epoch": 2.66,
      "learning_rate": 2.3691311386940098e-05,
      "loss": 0.7971,
      "step": 20000
    },
    {
      "epoch": 2.73,
      "learning_rate": 2.301672962763087e-05,
      "loss": 0.8096,
      "step": 20500
    },
    {
      "epoch": 2.8,
      "learning_rate": 2.2342147868321642e-05,
      "loss": 0.7909,
      "step": 21000
    },
    {
      "epoch": 2.86,
      "learning_rate": 2.1667566109012415e-05,
      "loss": 0.7939,
      "step": 21500
    },
    {
      "epoch": 2.93,
      "learning_rate": 2.0992984349703183e-05,
      "loss": 0.7963,
      "step": 22000
    },
    {
      "epoch": 3.0,
      "learning_rate": 2.0318402590393955e-05,
      "loss": 0.7886,
      "step": 22500
    },
    {
      "epoch": 3.0,
      "eval_bleu1_score": 0.9978,
      "eval_bleu2_score": 0.9978,
      "eval_bleu3_score": 0.9978,
      "eval_bleu4_score": 0.9978,
      "eval_loss": 1.1093112230300903,
      "eval_rougeL": 0.9976,
      "eval_runtime": 347.4161,
      "eval_samples_per_second": 4.807,
      "eval_steps_per_second": 2.403,
      "step": 22536
    },
    {
      "epoch": 3.06,
      "learning_rate": 1.9643820831084728e-05,
      "loss": 0.576,
      "step": 23000
    },
    {
      "epoch": 3.13,
      "learning_rate": 1.89692390717755e-05,
      "loss": 0.5781,
      "step": 23500
    },
    {
      "epoch": 3.19,
      "learning_rate": 1.8294657312466272e-05,
      "loss": 0.5698,
      "step": 24000
    },
    {
      "epoch": 3.26,
      "learning_rate": 1.762007555315704e-05,
      "loss": 0.5717,
      "step": 24500
    },
    {
      "epoch": 3.33,
      "learning_rate": 1.6945493793847816e-05,
      "loss": 0.584,
      "step": 25000
    },
    {
      "epoch": 3.39,
      "learning_rate": 1.6270912034538588e-05,
      "loss": 0.5758,
      "step": 25500
    },
    {
      "epoch": 3.46,
      "learning_rate": 1.559633027522936e-05,
      "loss": 0.5808,
      "step": 26000
    },
    {
      "epoch": 3.53,
      "learning_rate": 1.492174851592013e-05,
      "loss": 0.5762,
      "step": 26500
    },
    {
      "epoch": 3.59,
      "learning_rate": 1.4247166756610903e-05,
      "loss": 0.5701,
      "step": 27000
    },
    {
      "epoch": 3.66,
      "learning_rate": 1.3572584997301673e-05,
      "loss": 0.58,
      "step": 27500
    },
    {
      "epoch": 3.73,
      "learning_rate": 1.2898003237992445e-05,
      "loss": 0.5683,
      "step": 28000
    },
    {
      "epoch": 3.79,
      "learning_rate": 1.2223421478683218e-05,
      "loss": 0.5589,
      "step": 28500
    },
    {
      "epoch": 3.86,
      "learning_rate": 1.1548839719373988e-05,
      "loss": 0.5842,
      "step": 29000
    },
    {
      "epoch": 3.93,
      "learning_rate": 1.087425796006476e-05,
      "loss": 0.5717,
      "step": 29500
    },
    {
      "epoch": 3.99,
      "learning_rate": 1.0199676200755532e-05,
      "loss": 0.5618,
      "step": 30000
    },
    {
      "epoch": 4.0,
      "eval_bleu1_score": 0.9978,
      "eval_bleu2_score": 0.9978,
      "eval_bleu3_score": 0.9978,
      "eval_bleu4_score": 0.9978,
      "eval_loss": 1.2060836553573608,
      "eval_rougeL": 0.9976,
      "eval_runtime": 347.0486,
      "eval_samples_per_second": 4.812,
      "eval_steps_per_second": 2.406,
      "step": 30048
    },
    {
      "epoch": 4.06,
      "learning_rate": 9.525094441446304e-06,
      "loss": 0.3953,
      "step": 30500
    },
    {
      "epoch": 4.13,
      "learning_rate": 8.850512682137076e-06,
      "loss": 0.3749,
      "step": 31000
    },
    {
      "epoch": 4.19,
      "learning_rate": 8.175930922827847e-06,
      "loss": 0.3731,
      "step": 31500
    },
    {
      "epoch": 4.26,
      "learning_rate": 7.501349163518619e-06,
      "loss": 0.3752,
      "step": 32000
    },
    {
      "epoch": 4.33,
      "learning_rate": 6.82676740420939e-06,
      "loss": 0.3789,
      "step": 32500
    },
    {
      "epoch": 4.39,
      "learning_rate": 6.152185644900162e-06,
      "loss": 0.3768,
      "step": 33000
    },
    {
      "epoch": 4.46,
      "learning_rate": 5.477603885590934e-06,
      "loss": 0.3665,
      "step": 33500
    },
    {
      "epoch": 4.53,
      "learning_rate": 4.803022126281705e-06,
      "loss": 0.3763,
      "step": 34000
    },
    {
      "epoch": 4.59,
      "learning_rate": 4.128440366972477e-06,
      "loss": 0.3715,
      "step": 34500
    },
    {
      "epoch": 4.66,
      "learning_rate": 3.4538586076632488e-06,
      "loss": 0.3763,
      "step": 35000
    },
    {
      "epoch": 4.73,
      "learning_rate": 2.7792768483540205e-06,
      "loss": 0.3727,
      "step": 35500
    },
    {
      "epoch": 4.79,
      "learning_rate": 2.104695089044792e-06,
      "loss": 0.365,
      "step": 36000
    },
    {
      "epoch": 4.86,
      "learning_rate": 1.4301133297355639e-06,
      "loss": 0.3707,
      "step": 36500
    },
    {
      "epoch": 4.93,
      "learning_rate": 7.555315704263357e-07,
      "loss": 0.3691,
      "step": 37000
    },
    {
      "epoch": 4.99,
      "learning_rate": 8.09498111171074e-08,
      "loss": 0.3724,
      "step": 37500
    },
    {
      "epoch": 5.0,
      "eval_bleu1_score": 0.9978,
      "eval_bleu2_score": 0.9978,
      "eval_bleu3_score": 0.9978,
      "eval_bleu4_score": 0.9978,
      "eval_loss": 1.51046884059906,
      "eval_rougeL": 0.9976,
      "eval_runtime": 347.5328,
      "eval_samples_per_second": 4.805,
      "eval_steps_per_second": 2.403,
      "step": 37560
    }
  ],
  "max_steps": 37560,
  "num_train_epochs": 5,
  "total_flos": 1.08999726956544e+17,
  "trial_name": null,
  "trial_params": null
}