trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 1.995967741935484,
  "eval_steps": 495,
  "global_step": 990,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.2,
      "grad_norm": 0.7123413681983948,
      "learning_rate": 0.0002,
      "loss": 0.9903,
      "step": 100
    },
    {
      "epoch": 0.4,
      "grad_norm": 0.5445395112037659,
      "learning_rate": 0.0002,
      "loss": 0.5847,
      "step": 200
    },
    {
      "epoch": 0.6,
      "grad_norm": 0.5527735948562622,
      "learning_rate": 0.0002,
      "loss": 0.4054,
      "step": 300
    },
    {
      "epoch": 0.81,
      "grad_norm": 0.3411148190498352,
      "learning_rate": 0.0002,
      "loss": 0.2877,
      "step": 400
    },
    {
      "epoch": 1.0,
      "eval_loss": 0.1955450177192688,
      "eval_runtime": 459.2044,
      "eval_samples_per_second": 7.681,
      "eval_steps_per_second": 1.921,
      "step": 495
    },
    {
      "epoch": 1.01,
      "grad_norm": 0.21418581902980804,
      "learning_rate": 0.0002,
      "loss": 0.215,
      "step": 500
    },
    {
      "epoch": 1.21,
      "grad_norm": 0.2143397480249405,
      "learning_rate": 0.0002,
      "loss": 0.1507,
      "step": 600
    },
    {
      "epoch": 1.41,
      "grad_norm": 0.20695798099040985,
      "learning_rate": 0.0002,
      "loss": 0.1356,
      "step": 700
    },
    {
      "epoch": 1.61,
      "grad_norm": 0.1783648580312729,
      "learning_rate": 0.0002,
      "loss": 0.123,
      "step": 800
    },
    {
      "epoch": 1.81,
      "grad_norm": 0.1991637796163559,
      "learning_rate": 0.0002,
      "loss": 0.1129,
      "step": 900
    },
    {
      "epoch": 2.0,
      "eval_loss": 0.11679410934448242,
      "eval_runtime": 436.9065,
      "eval_samples_per_second": 8.073,
      "eval_steps_per_second": 2.019,
      "step": 990
    }
  ],
  "logging_steps": 100,
  "max_steps": 1000,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 3,
  "save_steps": 495,
  "total_flos": 1.2763485559860756e+18,
  "train_batch_size": 16,
  "trial_name": null,
  "trial_params": null
}