checkpoint-138/trainer_state.json

{
  "best_metric": 1.124284267425537,
  "best_model_checkpoint": "outputs/checkpoint-138",
  "epoch": 2.979757085020243,
  "eval_steps": 500,
  "global_step": 138,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.4318488529014845,
      "grad_norm": 1.7280967235565186,
      "learning_rate": 4e-05,
      "loss": 2.827,
      "step": 20
    },
    {
      "epoch": 0.863697705802969,
      "grad_norm": 0.49127423763275146,
      "learning_rate": 8e-05,
      "loss": 2.1949,
      "step": 40
    },
    {
      "epoch": 0.9932523616734144,
      "eval_loss": 1.7903296947479248,
      "eval_runtime": 12.0868,
      "eval_samples_per_second": 30.777,
      "eval_steps_per_second": 3.889,
      "step": 46
    },
    {
      "epoch": 1.2955465587044535,
      "grad_norm": 0.4634649455547333,
      "learning_rate": 0.00012,
      "loss": 1.9724,
      "step": 60
    },
    {
      "epoch": 1.7273954116059378,
      "grad_norm": 0.690120279788971,
      "learning_rate": 0.00016,
      "loss": 1.7592,
      "step": 80
    },
    {
      "epoch": 1.9865047233468287,
      "eval_loss": 1.3832396268844604,
      "eval_runtime": 12.0989,
      "eval_samples_per_second": 30.747,
      "eval_steps_per_second": 3.885,
      "step": 92
    },
    {
      "epoch": 2.1592442645074224,
      "grad_norm": 0.9806725382804871,
      "learning_rate": 0.0002,
      "loss": 1.6455,
      "step": 100
    },
    {
      "epoch": 2.591093117408907,
      "grad_norm": 0.978474497795105,
      "learning_rate": 0.00019943348002101371,
      "loss": 1.4659,
      "step": 120
    },
    {
      "epoch": 2.979757085020243,
      "eval_loss": 1.124284267425537,
      "eval_runtime": 12.1066,
      "eval_samples_per_second": 30.727,
      "eval_steps_per_second": 3.882,
      "step": 138
    }
  ],
  "logging_steps": 20,
  "max_steps": 690,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 15,
  "save_steps": 500,
  "total_flos": 1.2269163484299264e+16,
  "train_batch_size": 2,
  "trial_name": null,
  "trial_params": null
}