{
  "best_metric": 43.218,
  "best_model_checkpoint": "opus_big_adapt_wce_gloss_unsampled_precision_2_ubweight_1.5/checkpoint-80000",
  "epoch": 1.9716574245224892,
  "global_step": 80000,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.2,
      "learning_rate": 1.9753635243376465e-05,
      "loss": 0.1832,
      "step": 8000
    },
    {
      "epoch": 0.2,
      "eval_bleu": 42.311,
      "eval_gen_len": 33.3634,
      "eval_loss": 0.0981452614068985,
      "eval_runtime": 129.4343,
      "eval_samples_per_second": 8.058,
      "eval_steps_per_second": 0.51,
      "step": 8000
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.950723967960567e-05,
      "loss": 0.152,
      "step": 16000
    },
    {
      "epoch": 0.39,
      "eval_bleu": 42.5149,
      "eval_gen_len": 33.8552,
      "eval_loss": 0.09721900522708893,
      "eval_runtime": 121.8178,
      "eval_samples_per_second": 8.562,
      "eval_steps_per_second": 0.542,
      "step": 16000
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.9260844115834877e-05,
      "loss": 0.1477,
      "step": 24000
    },
    {
      "epoch": 0.59,
      "eval_bleu": 42.3053,
      "eval_gen_len": 33.1083,
      "eval_loss": 0.09773427993059158,
      "eval_runtime": 113.6446,
      "eval_samples_per_second": 9.178,
      "eval_steps_per_second": 0.581,
      "step": 24000
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.901447935921134e-05,
      "loss": 0.1437,
      "step": 32000
    },
    {
      "epoch": 0.79,
      "eval_bleu": 42.7689,
      "eval_gen_len": 34.2637,
      "eval_loss": 0.09667054563760757,
      "eval_runtime": 127.749,
      "eval_samples_per_second": 8.164,
      "eval_steps_per_second": 0.517,
      "step": 32000
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.8768114602587803e-05,
      "loss": 0.1422,
      "step": 40000
    },
    {
      "epoch": 0.99,
      "eval_bleu": 42.1023,
      "eval_gen_len": 34.2656,
      "eval_loss": 0.09601164609193802,
      "eval_runtime": 127.2919,
      "eval_samples_per_second": 8.194,
      "eval_steps_per_second": 0.518,
      "step": 40000
    },
    {
      "epoch": 1.18,
      "learning_rate": 1.8521749845964266e-05,
      "loss": 0.1265,
      "step": 48000
    },
    {
      "epoch": 1.18,
      "eval_bleu": 42.4637,
      "eval_gen_len": 34.1064,
      "eval_loss": 0.09702879190444946,
      "eval_runtime": 125.1346,
      "eval_samples_per_second": 8.335,
      "eval_steps_per_second": 0.527,
      "step": 48000
    },
    {
      "epoch": 1.38,
      "learning_rate": 1.8275415896487988e-05,
      "loss": 0.1269,
      "step": 56000
    },
    {
      "epoch": 1.38,
      "eval_bleu": 42.9093,
      "eval_gen_len": 33.7507,
      "eval_loss": 0.09716643393039703,
      "eval_runtime": 127.4089,
      "eval_samples_per_second": 8.186,
      "eval_steps_per_second": 0.518,
      "step": 56000
    },
    {
      "epoch": 1.58,
      "learning_rate": 1.8029020332717192e-05,
      "loss": 0.1275,
      "step": 64000
    },
    {
      "epoch": 1.58,
      "eval_bleu": 42.6221,
      "eval_gen_len": 34.4372,
      "eval_loss": 0.09658621996641159,
      "eval_runtime": 249.3691,
      "eval_samples_per_second": 4.183,
      "eval_steps_per_second": 0.265,
      "step": 64000
    },
    {
      "epoch": 1.77,
      "learning_rate": 1.7782655576093655e-05,
      "loss": 0.127,
      "step": 72000
    },
    {
      "epoch": 1.77,
      "eval_bleu": 42.3284,
      "eval_gen_len": 34.0441,
      "eval_loss": 0.09654054045677185,
      "eval_runtime": 197.536,
      "eval_samples_per_second": 5.28,
      "eval_steps_per_second": 0.334,
      "step": 72000
    },
    {
      "epoch": 1.97,
      "learning_rate": 1.7536321626617377e-05,
      "loss": 0.1266,
      "step": 80000
    },
    {
      "epoch": 1.97,
      "eval_bleu": 43.218,
      "eval_gen_len": 34.1035,
      "eval_loss": 0.09614226967096329,
      "eval_runtime": 124.4592,
      "eval_samples_per_second": 8.38,
      "eval_steps_per_second": 0.53,
      "step": 80000
    }
  ],
  "max_steps": 649200,
  "num_train_epochs": 16,
  "total_flos": 2.5844916958632346e+17,
  "trial_name": null,
  "trial_params": null
}