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": 5.773237921062508,
      "learning_rate": 8.333333333333333e-08,
      "logits/chosen": -2.801664352416992,
      "logits/rejected": -2.8308780193328857,
      "logps/chosen": -137.20802307128906,
      "logps/pi_response": -133.3135223388672,
      "logps/ref_response": -133.3135223388672,
      "logps/rejected": -141.6568145751953,
      "loss": 0.6933,
      "rewards/accuracies": 0.0,
      "rewards/chosen": 0.0,
      "rewards/margins": 0.0,
      "rewards/rejected": 0.0,
      "step": 1
    },
    {
      "epoch": 0.16736401673640167,
      "grad_norm": 5.901719507030408,
      "learning_rate": 4.930057285201027e-07,
      "logits/chosen": -2.7964792251586914,
      "logits/rejected": -2.7925798892974854,
      "logps/chosen": -155.59413146972656,
      "logps/pi_response": -153.61138916015625,
      "logps/ref_response": -153.69000244140625,
      "logps/rejected": -155.34437561035156,
      "loss": 0.6932,
      "rewards/accuracies": 0.4444444477558136,
      "rewards/chosen": 0.00010709436173783615,
      "rewards/margins": -0.00010032434511231259,
      "rewards/rejected": 0.0002074186922982335,
      "step": 10
    },
    {
      "epoch": 0.33472803347280333,
      "grad_norm": 5.900913266829362,
      "learning_rate": 4.187457503795526e-07,
      "logits/chosen": -2.8278727531433105,
      "logits/rejected": -2.815530300140381,
      "logps/chosen": -150.38006591796875,
      "logps/pi_response": -150.15914916992188,
      "logps/ref_response": -147.89337158203125,
      "logps/rejected": -152.52603149414062,
      "loss": 0.6936,
      "rewards/accuracies": 0.47187501192092896,
      "rewards/chosen": -0.023533424362540245,
      "rewards/margins": -0.0003430193173699081,
      "rewards/rejected": -0.023190408945083618,
      "step": 20
    },
    {
      "epoch": 0.502092050209205,
      "grad_norm": 6.043686716544743,
      "learning_rate": 2.8691164100062034e-07,
      "logits/chosen": -2.7830755710601807,
      "logits/rejected": -2.7708182334899902,
      "logps/chosen": -158.68783569335938,
      "logps/pi_response": -158.1934814453125,
      "logps/ref_response": -141.50277709960938,
      "logps/rejected": -158.74456787109375,
      "loss": 0.6931,
      "rewards/accuracies": 0.5,
      "rewards/chosen": -0.16305488348007202,
      "rewards/margins": 0.0035704367328435183,
      "rewards/rejected": -0.16662533581256866,
      "step": 30
    },
    {
      "epoch": 0.6694560669456067,
      "grad_norm": 6.407268964437461,
      "learning_rate": 1.4248369943086995e-07,
      "logits/chosen": -2.7545723915100098,
      "logits/rejected": -2.753416061401367,
      "logps/chosen": -155.69338989257812,
      "logps/pi_response": -155.184814453125,
      "logps/ref_response": -143.01797485351562,
      "logps/rejected": -154.00015258789062,
      "loss": 0.6917,
      "rewards/accuracies": 0.4906249940395355,
      "rewards/chosen": -0.1169251948595047,
      "rewards/margins": 0.006899774074554443,
      "rewards/rejected": -0.12382496893405914,
      "step": 40
    },
    {
      "epoch": 0.8368200836820083,
      "grad_norm": 6.008222457230155,
      "learning_rate": 3.473909705816111e-08,
      "logits/chosen": -2.7677838802337646,
      "logits/rejected": -2.7679364681243896,
      "logps/chosen": -142.987060546875,
      "logps/pi_response": -143.57041931152344,
      "logps/ref_response": -138.1813201904297,
      "logps/rejected": -140.4540557861328,
      "loss": 0.6914,
      "rewards/accuracies": 0.534375011920929,
      "rewards/chosen": -0.053286753594875336,
      "rewards/margins": 0.004001016262918711,
      "rewards/rejected": -0.05728777125477791,
      "step": 50
    },
    {
      "epoch": 0.9874476987447699,
      "step": 59,
      "total_flos": 0.0,
      "train_loss": 0.6922930054745432,
      "train_runtime": 2626.0315,
      "train_samples_per_second": 5.82,
      "train_steps_per_second": 0.022
    }
  ],
  "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": 4,
  "trial_name": null,
  "trial_params": null
}