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": 58.266818387163894,
      "learning_rate": 8.333333333333333e-08,
      "logits/chosen": -2.640246868133545,
      "logits/rejected": -2.6412415504455566,
      "logps/chosen": -221.59217834472656,
      "logps/pi_response": -146.05226135253906,
      "logps/ref_response": -146.05226135253906,
      "logps/rejected": -459.28009033203125,
      "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": 41.65988474449752,
      "learning_rate": 4.930057285201027e-07,
      "logits/chosen": -2.632171630859375,
      "logits/rejected": -2.585049629211426,
      "logps/chosen": -243.53781127929688,
      "logps/pi_response": -126.54959106445312,
      "logps/ref_response": -117.57322692871094,
      "logps/rejected": -443.431884765625,
      "loss": 0.6436,
      "rewards/accuracies": 0.5868055820465088,
      "rewards/chosen": -0.20465856790542603,
      "rewards/margins": 0.1644359976053238,
      "rewards/rejected": -0.3690946102142334,
      "step": 10
    },
    {
      "epoch": 0.33472803347280333,
      "grad_norm": 40.52264024295099,
      "learning_rate": 4.187457503795526e-07,
      "logits/chosen": -2.519150972366333,
      "logits/rejected": -2.4623703956604004,
      "logps/chosen": -364.3970031738281,
      "logps/pi_response": -168.98965454101562,
      "logps/ref_response": -125.69271087646484,
      "logps/rejected": -663.7247314453125,
      "loss": 0.5606,
      "rewards/accuracies": 0.75,
      "rewards/chosen": -1.2418674230575562,
      "rewards/margins": 1.2894560098648071,
      "rewards/rejected": -2.5313236713409424,
      "step": 20
    },
    {
      "epoch": 0.502092050209205,
      "grad_norm": 27.07562924150943,
      "learning_rate": 2.8691164100062034e-07,
      "logits/chosen": -2.4185967445373535,
      "logits/rejected": -2.362191677093506,
      "logps/chosen": -338.4853820800781,
      "logps/pi_response": -144.1025848388672,
      "logps/ref_response": -121.75166320800781,
      "logps/rejected": -581.2408447265625,
      "loss": 0.4885,
      "rewards/accuracies": 0.7406250238418579,
      "rewards/chosen": -0.8758844137191772,
      "rewards/margins": 1.0427850484848022,
      "rewards/rejected": -1.9186694622039795,
      "step": 30
    },
    {
      "epoch": 0.6694560669456067,
      "grad_norm": 29.889956038106074,
      "learning_rate": 1.4248369943086995e-07,
      "logits/chosen": -2.4027583599090576,
      "logits/rejected": -2.3214104175567627,
      "logps/chosen": -301.30023193359375,
      "logps/pi_response": -138.3297576904297,
      "logps/ref_response": -122.65152740478516,
      "logps/rejected": -591.5069580078125,
      "loss": 0.4636,
      "rewards/accuracies": 0.793749988079071,
      "rewards/chosen": -0.6572309136390686,
      "rewards/margins": 1.041191816329956,
      "rewards/rejected": -1.6984226703643799,
      "step": 40
    },
    {
      "epoch": 0.8368200836820083,
      "grad_norm": 32.7260861871924,
      "learning_rate": 3.473909705816111e-08,
      "logits/chosen": -2.377892255783081,
      "logits/rejected": -2.300891160964966,
      "logps/chosen": -314.6452331542969,
      "logps/pi_response": -141.83682250976562,
      "logps/ref_response": -123.07878112792969,
      "logps/rejected": -590.2091674804688,
      "loss": 0.4642,
      "rewards/accuracies": 0.784375011920929,
      "rewards/chosen": -0.699453592300415,
      "rewards/margins": 1.0583473443984985,
      "rewards/rejected": -1.7578010559082031,
      "step": 50
    },
    {
      "epoch": 0.9874476987447699,
      "step": 59,
      "total_flos": 0.0,
      "train_loss": 0.5172020217119637,
      "train_runtime": 2584.7224,
      "train_samples_per_second": 5.913,
      "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
}