trainer_state.json

{
  "best_global_step": 6250,
  "best_metric": 59.5126771383472,
  "best_model_checkpoint": "nllb-200-600M-dzo-eng-checkpoints/checkpoint-6250",
  "epoch": 5.0,
  "eval_steps": 500,
  "global_step": 6250,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.4,
      "grad_norm": 0.19286614656448364,
      "learning_rate": 2.76048e-05,
      "loss": 0.0794,
      "step": 500
    },
    {
      "epoch": 0.8,
      "grad_norm": 0.27002865076065063,
      "learning_rate": 2.52048e-05,
      "loss": 0.0765,
      "step": 1000
    },
    {
      "epoch": 1.0,
      "eval_bleu": 58.03728575479501,
      "eval_loss": 0.07460231333971024,
      "eval_runtime": 167.4219,
      "eval_samples_per_second": 5.973,
      "eval_steps_per_second": 1.493,
      "step": 1250
    },
    {
      "epoch": 1.2,
      "grad_norm": 0.3723973035812378,
      "learning_rate": 2.28048e-05,
      "loss": 0.0644,
      "step": 1500
    },
    {
      "epoch": 1.6,
      "grad_norm": 0.20845282077789307,
      "learning_rate": 2.04048e-05,
      "loss": 0.0576,
      "step": 2000
    },
    {
      "epoch": 2.0,
      "grad_norm": 0.3116896450519562,
      "learning_rate": 1.80048e-05,
      "loss": 0.0576,
      "step": 2500
    },
    {
      "epoch": 2.0,
      "eval_bleu": 58.57463505915474,
      "eval_loss": 0.07282423228025436,
      "eval_runtime": 168.9336,
      "eval_samples_per_second": 5.919,
      "eval_steps_per_second": 1.48,
      "step": 2500
    },
    {
      "epoch": 2.4,
      "grad_norm": 0.2609412372112274,
      "learning_rate": 1.56048e-05,
      "loss": 0.0452,
      "step": 3000
    },
    {
      "epoch": 2.8,
      "grad_norm": 0.16577185690402985,
      "learning_rate": 1.3204800000000001e-05,
      "loss": 0.0465,
      "step": 3500
    },
    {
      "epoch": 3.0,
      "eval_bleu": 59.30988278861181,
      "eval_loss": 0.07350268214941025,
      "eval_runtime": 171.8963,
      "eval_samples_per_second": 5.817,
      "eval_steps_per_second": 1.454,
      "step": 3750
    },
    {
      "epoch": 3.2,
      "grad_norm": 0.1930246204137802,
      "learning_rate": 1.08048e-05,
      "loss": 0.0426,
      "step": 4000
    },
    {
      "epoch": 3.6,
      "grad_norm": 0.3314656913280487,
      "learning_rate": 8.404800000000001e-06,
      "loss": 0.0381,
      "step": 4500
    },
    {
      "epoch": 4.0,
      "grad_norm": 0.26522204279899597,
      "learning_rate": 6.0048000000000005e-06,
      "loss": 0.0381,
      "step": 5000
    },
    {
      "epoch": 4.0,
      "eval_bleu": 59.249316493031536,
      "eval_loss": 0.07584889233112335,
      "eval_runtime": 168.0339,
      "eval_samples_per_second": 5.951,
      "eval_steps_per_second": 1.488,
      "step": 5000
    },
    {
      "epoch": 4.4,
      "grad_norm": 0.17317089438438416,
      "learning_rate": 3.6048e-06,
      "loss": 0.0337,
      "step": 5500
    },
    {
      "epoch": 4.8,
      "grad_norm": 0.32834669947624207,
      "learning_rate": 1.2048e-06,
      "loss": 0.033,
      "step": 6000
    },
    {
      "epoch": 5.0,
      "eval_bleu": 59.5126771383472,
      "eval_loss": 0.07744310051202774,
      "eval_runtime": 168.6113,
      "eval_samples_per_second": 5.931,
      "eval_steps_per_second": 1.483,
      "step": 6250
    },
    {
      "epoch": 5.0,
      "step": 6250,
      "total_flos": 1.35444037632e+16,
      "train_loss": 0.05036833358764648,
      "train_runtime": 6278.8779,
      "train_samples_per_second": 7.963,
      "train_steps_per_second": 0.995
    }
  ],
  "logging_steps": 500,
  "max_steps": 6250,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 5,
  "save_steps": 500,
  "stateful_callbacks": {
    "TrainerControl": {
      "args": {
        "should_epoch_stop": false,
        "should_evaluate": false,
        "should_log": false,
        "should_save": true,
        "should_training_stop": true
      },
      "attributes": {}
    }
  },
  "total_flos": 1.35444037632e+16,
  "train_batch_size": 4,
  "trial_name": null,
  "trial_params": null
}