last-checkpoint/trainer_state.json

{
  "best_metric": 1.4787811040878296,
  "best_model_checkpoint": "./outputs/checkpoint-500",
  "epoch": 0.6666666666666666,
  "eval_steps": 100,
  "global_step": 500,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.13,
      "learning_rate": 0.0002,
      "loss": 1.4578,
      "step": 100
    },
    {
      "epoch": 0.13,
      "eval_loss": 1.567879319190979,
      "eval_runtime": 93.4626,
      "eval_samples_per_second": 15.525,
      "eval_steps_per_second": 1.947,
      "step": 100
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0002,
      "loss": 1.2977,
      "step": 200
    },
    {
      "epoch": 0.27,
      "eval_loss": 1.5340639352798462,
      "eval_runtime": 93.3831,
      "eval_samples_per_second": 15.538,
      "eval_steps_per_second": 1.949,
      "step": 200
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0002,
      "loss": 1.2762,
      "step": 300
    },
    {
      "epoch": 0.4,
      "eval_loss": 1.5129507780075073,
      "eval_runtime": 93.3176,
      "eval_samples_per_second": 15.549,
      "eval_steps_per_second": 1.95,
      "step": 300
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0002,
      "loss": 1.2673,
      "step": 400
    },
    {
      "epoch": 0.53,
      "eval_loss": 1.4979034662246704,
      "eval_runtime": 93.4276,
      "eval_samples_per_second": 15.531,
      "eval_steps_per_second": 1.948,
      "step": 400
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0002,
      "loss": 1.2497,
      "step": 500
    },
    {
      "epoch": 0.67,
      "eval_loss": 1.4787811040878296,
      "eval_runtime": 93.1297,
      "eval_samples_per_second": 15.58,
      "eval_steps_per_second": 1.954,
      "step": 500
    }
  ],
  "logging_steps": 100,
  "max_steps": 2250,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 3,
  "save_steps": 100,
  "total_flos": 2.927413222981632e+16,
  "trial_name": null,
  "trial_params": null
}