last-checkpoint/trainer_state.json

{
  "best_metric": 1.1881097555160522,
  "best_model_checkpoint": "../facebook/nllb-200-3.3B-finetuned/checkpoint-1000",
  "epoch": 2000.0,
  "eval_steps": 500,
  "global_step": 2000,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 500.0,
      "grad_norm": 13.488256454467773,
      "learning_rate": 9.758186548415274e-06,
      "loss": 4.34,
      "step": 500
    },
    {
      "epoch": 500.0,
      "eval_gen_len": 26.0,
      "eval_loss": 2.901174306869507,
      "eval_rouge": 0.05,
      "eval_runtime": 7.0444,
      "eval_samples_per_second": 0.284,
      "eval_steps_per_second": 0.142,
      "step": 500
    },
    {
      "epoch": 1000.0,
      "grad_norm": 0.9894864559173584,
      "learning_rate": 9.050617527877911e-06,
      "loss": 0.609,
      "step": 1000
    },
    {
      "epoch": 1000.0,
      "eval_gen_len": 26.0,
      "eval_loss": 1.1881097555160522,
      "eval_rouge": 0.098,
      "eval_runtime": 1.1556,
      "eval_samples_per_second": 1.731,
      "eval_steps_per_second": 0.865,
      "step": 1000
    },
    {
      "epoch": 1500.0,
      "grad_norm": 0.1642373949289322,
      "learning_rate": 7.946545841400035e-06,
      "loss": 0.0196,
      "step": 1500
    },
    {
      "epoch": 1500.0,
      "eval_gen_len": 30.5,
      "eval_loss": 1.5324629545211792,
      "eval_rouge": 0.1493,
      "eval_runtime": 6.9442,
      "eval_samples_per_second": 0.288,
      "eval_steps_per_second": 0.144,
      "step": 1500
    },
    {
      "epoch": 2000.0,
      "grad_norm": 0.08364015817642212,
      "learning_rate": 6.554045718173867e-06,
      "loss": 0.0061,
      "step": 2000
    },
    {
      "epoch": 2000.0,
      "eval_gen_len": 33.5,
      "eval_loss": 1.5447587966918945,
      "eval_rouge": 0.125,
      "eval_runtime": 6.9455,
      "eval_samples_per_second": 0.288,
      "eval_steps_per_second": 0.144,
      "step": 2000
    }
  ],
  "logging_steps": 500,
  "max_steps": 5000,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 5000,
  "save_steps": 500,
  "total_flos": 1.2724531888128e+16,
  "train_batch_size": 8,
  "trial_name": null,
  "trial_params": null
}