trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 2.769230769230769,
  "eval_steps": 500,
  "global_step": 27,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.1,
      "grad_norm": 6.454160266082772,
      "learning_rate": 6.666666666666667e-06,
      "loss": 0.6068,
      "step": 1
    },
    {
      "epoch": 0.51,
      "grad_norm": 23.663284070156248,
      "learning_rate": 1.9659258262890683e-05,
      "loss": 0.7615,
      "step": 5
    },
    {
      "epoch": 0.92,
      "eval_loss": 0.8382714986801147,
      "eval_runtime": 14.2719,
      "eval_samples_per_second": 8.058,
      "eval_steps_per_second": 0.561,
      "step": 9
    },
    {
      "epoch": 1.03,
      "grad_norm": 4.052343653228325,
      "learning_rate": 1.608761429008721e-05,
      "loss": 0.7722,
      "step": 10
    },
    {
      "epoch": 1.54,
      "grad_norm": 2.102662884979,
      "learning_rate": 1e-05,
      "loss": 0.5005,
      "step": 15
    },
    {
      "epoch": 1.95,
      "eval_loss": 0.6564590334892273,
      "eval_runtime": 14.3846,
      "eval_samples_per_second": 7.995,
      "eval_steps_per_second": 0.556,
      "step": 19
    },
    {
      "epoch": 2.05,
      "grad_norm": 1.7730868971351998,
      "learning_rate": 3.912385709912794e-06,
      "loss": 0.4245,
      "step": 20
    },
    {
      "epoch": 2.56,
      "grad_norm": 1.2908369244318303,
      "learning_rate": 3.4074173710931804e-07,
      "loss": 0.365,
      "step": 25
    },
    {
      "epoch": 2.77,
      "eval_loss": 0.6345903873443604,
      "eval_runtime": 14.8474,
      "eval_samples_per_second": 7.745,
      "eval_steps_per_second": 0.539,
      "step": 27
    },
    {
      "epoch": 2.77,
      "step": 27,
      "total_flos": 5600905789440.0,
      "train_loss": 0.5430978823591162,
      "train_runtime": 627.6728,
      "train_samples_per_second": 2.92,
      "train_steps_per_second": 0.043
    }
  ],
  "logging_steps": 5,
  "max_steps": 27,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 3,
  "save_steps": 500,
  "total_flos": 5600905789440.0,
  "train_batch_size": 4,
  "trial_name": null,
  "trial_params": null
}