trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 0.9942857142857143,
  "eval_steps": 100,
  "global_step": 87,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.01,
      "grad_norm": 19.437784440370766,
      "learning_rate": 5.555555555555555e-08,
      "logits/chosen": -2.6343841552734375,
      "logits/rejected": -2.6980783939361572,
      "logps/chosen": -581.0560302734375,
      "logps/rejected": -816.3157958984375,
      "loss": 0.6931,
      "rewards/accuracies": 0.0,
      "rewards/chosen": 0.0,
      "rewards/margins": 0.0,
      "rewards/rejected": 0.0,
      "step": 1
    },
    {
      "epoch": 0.11,
      "grad_norm": 24.044765874363524,
      "learning_rate": 4.997972495428924e-07,
      "logits/chosen": -2.592595100402832,
      "logits/rejected": -2.6620125770568848,
      "logps/chosen": -575.6597900390625,
      "logps/rejected": -842.704345703125,
      "loss": 0.6913,
      "rewards/accuracies": 0.5277777910232544,
      "rewards/chosen": 0.00319318613037467,
      "rewards/margins": 0.0027422416023910046,
      "rewards/rejected": 0.00045094432425685227,
      "step": 10
    },
    {
      "epoch": 0.23,
      "grad_norm": 21.402086416348496,
      "learning_rate": 4.7586260865259554e-07,
      "logits/chosen": -2.666720151901245,
      "logits/rejected": -2.636192798614502,
      "logps/chosen": -614.2251586914062,
      "logps/rejected": -898.0855712890625,
      "loss": 0.6594,
      "rewards/accuracies": 0.7875000238418579,
      "rewards/chosen": 0.053952403366565704,
      "rewards/margins": 0.0747995600104332,
      "rewards/rejected": -0.020847156643867493,
      "step": 20
    },
    {
      "epoch": 0.34,
      "grad_norm": 30.86499685610669,
      "learning_rate": 4.157806645601988e-07,
      "logits/chosen": -2.8280863761901855,
      "logits/rejected": -2.7879080772399902,
      "logps/chosen": -507.06854248046875,
      "logps/rejected": -962.9939575195312,
      "loss": 0.5746,
      "rewards/accuracies": 0.7875000238418579,
      "rewards/chosen": -0.026104014366865158,
      "rewards/margins": 0.35351747274398804,
      "rewards/rejected": -0.3796215355396271,
      "step": 30
    },
    {
      "epoch": 0.46,
      "grad_norm": 47.705913880090655,
      "learning_rate": 3.2916699845036815e-07,
      "logits/chosen": -3.0843756198883057,
      "logits/rejected": -3.0327470302581787,
      "logps/chosen": -613.1222534179688,
      "logps/rejected": -1105.1820068359375,
      "loss": 0.4397,
      "rewards/accuracies": 0.8062499761581421,
      "rewards/chosen": -0.6070794463157654,
      "rewards/margins": 1.289810299873352,
      "rewards/rejected": -1.8968894481658936,
      "step": 40
    },
    {
      "epoch": 0.57,
      "grad_norm": 30.02360571000533,
      "learning_rate": 2.2988335782081851e-07,
      "logits/chosen": -2.9371981620788574,
      "logits/rejected": -3.0881218910217285,
      "logps/chosen": -528.0693359375,
      "logps/rejected": -1141.9866943359375,
      "loss": 0.378,
      "rewards/accuracies": 0.8687499761581421,
      "rewards/chosen": -0.27318429946899414,
      "rewards/margins": 1.5688350200653076,
      "rewards/rejected": -1.8420193195343018,
      "step": 50
    },
    {
      "epoch": 0.69,
      "grad_norm": 32.099661091341254,
      "learning_rate": 1.3381920698905784e-07,
      "logits/chosen": -2.9197936058044434,
      "logits/rejected": -3.0912833213806152,
      "logps/chosen": -550.3436279296875,
      "logps/rejected": -1202.0462646484375,
      "loss": 0.3308,
      "rewards/accuracies": 0.84375,
      "rewards/chosen": -0.4591130316257477,
      "rewards/margins": 2.357220411300659,
      "rewards/rejected": -2.816333532333374,
      "step": 60
    },
    {
      "epoch": 0.8,
      "grad_norm": 32.99216577780543,
      "learning_rate": 5.6348759543086374e-08,
      "logits/chosen": -2.9525704383850098,
      "logits/rejected": -3.2371833324432373,
      "logps/chosen": -646.6183471679688,
      "logps/rejected": -1339.00146484375,
      "loss": 0.3031,
      "rewards/accuracies": 0.9125000238418579,
      "rewards/chosen": -0.535283625125885,
      "rewards/margins": 2.8484416007995605,
      "rewards/rejected": -3.38372540473938,
      "step": 70
    },
    {
      "epoch": 0.91,
      "grad_norm": 36.54372644480357,
      "learning_rate": 9.87047209215694e-09,
      "logits/chosen": -2.80966854095459,
      "logits/rejected": -3.226341962814331,
      "logps/chosen": -611.0046997070312,
      "logps/rejected": -1294.22900390625,
      "loss": 0.3159,
      "rewards/accuracies": 0.8812500238418579,
      "rewards/chosen": -0.7080464363098145,
      "rewards/margins": 2.6834611892700195,
      "rewards/rejected": -3.391507625579834,
      "step": 80
    },
    {
      "epoch": 0.99,
      "step": 87,
      "total_flos": 0.0,
      "train_loss": 0.44661769921752226,
      "train_runtime": 1188.3148,
      "train_samples_per_second": 4.689,
      "train_steps_per_second": 0.073
    }
  ],
  "logging_steps": 10,
  "max_steps": 87,
  "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
}