trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 0.9984,
  "eval_steps": 500,
  "global_step": 156,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.32,
      "grad_norm": 4.15625,
      "learning_rate": 4.423334927457198e-05,
      "logits/chosen": 0.6726851463317871,
      "logits/rejected": 0.5370064973831177,
      "logps/chosen": -353.01397705078125,
      "logps/rejected": -321.87738037109375,
      "loss": 0.6045,
      "rewards/accuracies": 0.6434375047683716,
      "rewards/chosen": 0.3095881938934326,
      "rewards/margins": 0.49872708320617676,
      "rewards/rejected": -0.18913888931274414,
      "step": 50
    },
    {
      "epoch": 0.64,
      "grad_norm": 2.09375,
      "learning_rate": 1.815842524819793e-05,
      "logits/chosen": 0.6333445310592651,
      "logits/rejected": 0.49084267020225525,
      "logps/chosen": -334.806396484375,
      "logps/rejected": -316.8036804199219,
      "loss": 0.4899,
      "rewards/accuracies": 0.7724999785423279,
      "rewards/chosen": 0.6486607193946838,
      "rewards/margins": 1.2206343412399292,
      "rewards/rejected": -0.5719736814498901,
      "step": 100
    },
    {
      "epoch": 0.96,
      "grad_norm": 3.546875,
      "learning_rate": 2.397392281198729e-07,
      "logits/chosen": 0.6054231524467468,
      "logits/rejected": 0.45057016611099243,
      "logps/chosen": -348.30419921875,
      "logps/rejected": -318.5997314453125,
      "loss": 0.47,
      "rewards/accuracies": 0.7762500047683716,
      "rewards/chosen": 0.8248878717422485,
      "rewards/margins": 1.4257100820541382,
      "rewards/rejected": -0.6008222103118896,
      "step": 150
    },
    {
      "epoch": 0.9984,
      "step": 156,
      "total_flos": 4.792500052780122e+18,
      "train_loss": 0.5214443573584924,
      "train_runtime": 10667.9882,
      "train_samples_per_second": 0.937,
      "train_steps_per_second": 0.015
    }
  ],
  "logging_steps": 50,
  "max_steps": 156,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 1,
  "save_steps": 3906,
  "stateful_callbacks": {
    "TrainerControl": {
      "args": {
        "should_epoch_stop": false,
        "should_evaluate": false,
        "should_log": false,
        "should_save": true,
        "should_training_stop": true
      },
      "attributes": {}
    }
  },
  "total_flos": 4.792500052780122e+18,
  "train_batch_size": 1,
  "trial_name": null,
  "trial_params": null
}