trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 0.9874476987447699,
  "eval_steps": 500,
  "global_step": 59,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.016736401673640166,
      "grad_norm": 45.6561557696117,
      "learning_rate": 8.333333333333333e-08,
      "logits/chosen": -2.6417698860168457,
      "logits/rejected": -2.6596364974975586,
      "logps/chosen": -270.19268798828125,
      "logps/pi_response": -135.7864990234375,
      "logps/ref_response": -135.7864990234375,
      "logps/rejected": -406.362548828125,
      "loss": 0.6931,
      "rewards/accuracies": 0.0,
      "rewards/chosen": 0.0,
      "rewards/margins": 0.0,
      "rewards/rejected": 0.0,
      "step": 1
    },
    {
      "epoch": 0.16736401673640167,
      "grad_norm": 44.31989081570743,
      "learning_rate": 4.930057285201027e-07,
      "logits/chosen": -2.6331377029418945,
      "logits/rejected": -2.60306978225708,
      "logps/chosen": -258.35308837890625,
      "logps/pi_response": -124.52118682861328,
      "logps/ref_response": -115.27017974853516,
      "logps/rejected": -402.9174499511719,
      "loss": 0.6561,
      "rewards/accuracies": 0.6180555820465088,
      "rewards/chosen": -0.21156954765319824,
      "rewards/margins": 0.12729844450950623,
      "rewards/rejected": -0.33886802196502686,
      "step": 10
    },
    {
      "epoch": 0.33472803347280333,
      "grad_norm": 43.96985299144147,
      "learning_rate": 4.187457503795526e-07,
      "logits/chosen": -2.5259432792663574,
      "logits/rejected": -2.475297689437866,
      "logps/chosen": -358.1857604980469,
      "logps/pi_response": -152.20355224609375,
      "logps/ref_response": -118.438232421875,
      "logps/rejected": -570.7589111328125,
      "loss": 0.5793,
      "rewards/accuracies": 0.7124999761581421,
      "rewards/chosen": -1.0893189907073975,
      "rewards/margins": 1.0009855031967163,
      "rewards/rejected": -2.090304374694824,
      "step": 20
    },
    {
      "epoch": 0.502092050209205,
      "grad_norm": 26.005322215555676,
      "learning_rate": 2.8691164100062034e-07,
      "logits/chosen": -2.458430767059326,
      "logits/rejected": -2.390451192855835,
      "logps/chosen": -332.345703125,
      "logps/pi_response": -141.60523986816406,
      "logps/ref_response": -112.2582778930664,
      "logps/rejected": -509.4139709472656,
      "loss": 0.51,
      "rewards/accuracies": 0.7562500238418579,
      "rewards/chosen": -0.818560779094696,
      "rewards/margins": 0.7382112741470337,
      "rewards/rejected": -1.556771993637085,
      "step": 30
    },
    {
      "epoch": 0.6694560669456067,
      "grad_norm": 22.090907027333337,
      "learning_rate": 1.4248369943086995e-07,
      "logits/chosen": -2.44326114654541,
      "logits/rejected": -2.3678460121154785,
      "logps/chosen": -296.580810546875,
      "logps/pi_response": -139.00628662109375,
      "logps/ref_response": -118.343017578125,
      "logps/rejected": -528.0565185546875,
      "loss": 0.4894,
      "rewards/accuracies": 0.7718750238418579,
      "rewards/chosen": -0.6138304471969604,
      "rewards/margins": 0.8052016496658325,
      "rewards/rejected": -1.4190319776535034,
      "step": 40
    },
    {
      "epoch": 0.8368200836820083,
      "grad_norm": 27.68112719604131,
      "learning_rate": 3.473909705816111e-08,
      "logits/chosen": -2.3917486667633057,
      "logits/rejected": -2.3199546337127686,
      "logps/chosen": -333.00665283203125,
      "logps/pi_response": -144.41848754882812,
      "logps/ref_response": -117.71931457519531,
      "logps/rejected": -528.3931274414062,
      "loss": 0.4903,
      "rewards/accuracies": 0.731249988079071,
      "rewards/chosen": -0.8060160875320435,
      "rewards/margins": 0.7739337086677551,
      "rewards/rejected": -1.5799497365951538,
      "step": 50
    },
    {
      "epoch": 0.9874476987447699,
      "step": 59,
      "total_flos": 0.0,
      "train_loss": 0.5345395055867858,
      "train_runtime": 2589.1136,
      "train_samples_per_second": 5.903,
      "train_steps_per_second": 0.023
    }
  ],
  "logging_steps": 10,
  "max_steps": 59,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 1,
  "save_steps": 100,
  "total_flos": 0.0,
  "train_batch_size": 8,
  "trial_name": null,
  "trial_params": null
}