checkpoint-376/trainer_state.json

{
  "best_metric": 1.1512998342514038,
  "best_model_checkpoint": "./checkpoints/dpo-mix-7k/Qwen1.5-0.5B-dpo-mix-7k-lambda1.0-ORPO-29-9-20/checkpoint-376",
  "epoch": 2.0,
  "eval_steps": 500,
  "global_step": 376,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.27,
      "grad_norm": 676.0,
      "learning_rate": 2.5e-06,
      "loss": 15.3348,
      "step": 50
    },
    {
      "epoch": 0.53,
      "grad_norm": 16.875,
      "learning_rate": 5e-06,
      "loss": 3.911,
      "step": 100
    },
    {
      "epoch": 0.8,
      "grad_norm": 5.75,
      "learning_rate": 4.6059265481735295e-06,
      "loss": 1.1674,
      "step": 150
    },
    {
      "epoch": 1.0,
      "eval_loss": 1.157950758934021,
      "eval_runtime": 34.0907,
      "eval_samples_per_second": 16.573,
      "eval_steps_per_second": 1.056,
      "step": 188
    },
    {
      "epoch": 1.06,
      "grad_norm": 6.46875,
      "learning_rate": 3.5479413010416606e-06,
      "loss": 1.1863,
      "step": 200
    },
    {
      "epoch": 1.33,
      "grad_norm": 7.78125,
      "learning_rate": 2.159583377259384e-06,
      "loss": 1.1619,
      "step": 250
    },
    {
      "epoch": 1.6,
      "grad_norm": 5.46875,
      "learning_rate": 8.785447763431101e-07,
      "loss": 1.1292,
      "step": 300
    },
    {
      "epoch": 1.86,
      "grad_norm": 6.71875,
      "learning_rate": 1.0868414100166314e-07,
      "loss": 1.1423,
      "step": 350
    },
    {
      "epoch": 2.0,
      "eval_loss": 1.1512998342514038,
      "eval_runtime": 34.1097,
      "eval_samples_per_second": 16.564,
      "eval_steps_per_second": 1.055,
      "step": 376
    }
  ],
  "logging_steps": 50,
  "max_steps": 376,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 2,
  "save_steps": 500,
  "total_flos": 5699684433657856.0,
  "train_batch_size": 4,
  "trial_name": null,
  "trial_params": null
}