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": 84.02988583632249,
      "learning_rate": 1.6666666666666664e-08,
      "logits/chosen": -2.2970376014709473,
      "logits/rejected": -2.2431912422180176,
      "logps/chosen": -373.13104248046875,
      "logps/pi_response": -260.1656188964844,
      "logps/ref_response": -260.1656188964844,
      "logps/rejected": -692.877685546875,
      "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": 84.53092774390639,
      "learning_rate": 9.860114570402053e-08,
      "logits/chosen": -2.2980520725250244,
      "logits/rejected": -2.1980514526367188,
      "logps/chosen": -392.8827819824219,
      "logps/pi_response": -240.4452362060547,
      "logps/ref_response": -240.37266540527344,
      "logps/rejected": -701.8290405273438,
      "loss": 0.6838,
      "rewards/accuracies": 0.5833333134651184,
      "rewards/chosen": -0.010863525792956352,
      "rewards/margins": 0.021226655691862106,
      "rewards/rejected": -0.03209017589688301,
      "step": 10
    },
    {
      "epoch": 0.33472803347280333,
      "grad_norm": 51.47495174729241,
      "learning_rate": 8.374915007591053e-08,
      "logits/chosen": -2.2405943870544434,
      "logits/rejected": -2.166599750518799,
      "logps/chosen": -405.30322265625,
      "logps/pi_response": -246.46957397460938,
      "logps/ref_response": -241.376953125,
      "logps/rejected": -747.4080200195312,
      "loss": 0.5968,
      "rewards/accuracies": 0.7406250238418579,
      "rewards/chosen": -0.19080321490764618,
      "rewards/margins": 0.28054216504096985,
      "rewards/rejected": -0.47134536504745483,
      "step": 20
    },
    {
      "epoch": 0.502092050209205,
      "grad_norm": 32.300860121358355,
      "learning_rate": 5.738232820012406e-08,
      "logits/chosen": -2.169691562652588,
      "logits/rejected": -2.1060214042663574,
      "logps/chosen": -432.1522521972656,
      "logps/pi_response": -245.9633331298828,
      "logps/ref_response": -236.70889282226562,
      "logps/rejected": -788.1858520507812,
      "loss": 0.5361,
      "rewards/accuracies": 0.746874988079071,
      "rewards/chosen": -0.4271731972694397,
      "rewards/margins": 0.597862958908081,
      "rewards/rejected": -1.025036096572876,
      "step": 30
    },
    {
      "epoch": 0.6694560669456067,
      "grad_norm": 62.33777361652435,
      "learning_rate": 2.8496739886173992e-08,
      "logits/chosen": -2.153951644897461,
      "logits/rejected": -2.056622266769409,
      "logps/chosen": -465.5972595214844,
      "logps/pi_response": -267.73455810546875,
      "logps/ref_response": -249.88888549804688,
      "logps/rejected": -874.9528198242188,
      "loss": 0.5035,
      "rewards/accuracies": 0.7437499761581421,
      "rewards/chosen": -0.6645516157150269,
      "rewards/margins": 1.000557541847229,
      "rewards/rejected": -1.6651090383529663,
      "step": 40
    },
    {
      "epoch": 0.8368200836820083,
      "grad_norm": 108.35105270377402,
      "learning_rate": 6.947819411632222e-09,
      "logits/chosen": -2.1273913383483887,
      "logits/rejected": -2.047353982925415,
      "logps/chosen": -463.8194885253906,
      "logps/pi_response": -261.6892395019531,
      "logps/ref_response": -243.93222045898438,
      "logps/rejected": -947.5609130859375,
      "loss": 0.4965,
      "rewards/accuracies": 0.800000011920929,
      "rewards/chosen": -0.6754907965660095,
      "rewards/margins": 1.250502347946167,
      "rewards/rejected": -1.9259932041168213,
      "step": 50
    },
    {
      "epoch": 0.9874476987447699,
      "step": 59,
      "total_flos": 0.0,
      "train_loss": 0.549242844015865,
      "train_runtime": 2600.4955,
      "train_samples_per_second": 5.877,
      "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
}