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": 0.15945951840902217,
      "learning_rate": 8.333333333333333e-08,
      "logits/chosen": -1.923944354057312,
      "logits/rejected": -1.923944354057312,
      "logps/chosen": -152.47433471679688,
      "logps/pi_response": -152.48638916015625,
      "logps/ref_response": -152.48638916015625,
      "logps/rejected": -152.47433471679688,
      "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": 0.3170532004182247,
      "learning_rate": 4.930057285201027e-07,
      "logits/chosen": -1.9577691555023193,
      "logits/rejected": -1.9577691555023193,
      "logps/chosen": -169.63162231445312,
      "logps/pi_response": -169.6338653564453,
      "logps/ref_response": -151.2162322998047,
      "logps/rejected": -169.63162231445312,
      "loss": 0.6931,
      "rewards/accuracies": 0.0,
      "rewards/chosen": -0.18470749258995056,
      "rewards/margins": 0.0,
      "rewards/rejected": -0.18470749258995056,
      "step": 10
    },
    {
      "epoch": 0.33472803347280333,
      "grad_norm": 0.28020898661548793,
      "learning_rate": 4.187457503795526e-07,
      "logits/chosen": -1.840193510055542,
      "logits/rejected": -1.840193510055542,
      "logps/chosen": -359.313720703125,
      "logps/pi_response": -359.30316162109375,
      "logps/ref_response": -151.60671997070312,
      "logps/rejected": -359.313720703125,
      "loss": 0.6931,
      "rewards/accuracies": 0.0,
      "rewards/chosen": -2.077146053314209,
      "rewards/margins": 0.0,
      "rewards/rejected": -2.077146053314209,
      "step": 20
    },
    {
      "epoch": 0.502092050209205,
      "grad_norm": 0.3219085642544482,
      "learning_rate": 2.8691164100062034e-07,
      "logits/chosen": -1.2987468242645264,
      "logits/rejected": -1.2987468242645264,
      "logps/chosen": -588.2769165039062,
      "logps/pi_response": -588.268310546875,
      "logps/ref_response": -151.4405517578125,
      "logps/rejected": -588.2769165039062,
      "loss": 0.6931,
      "rewards/accuracies": 0.0,
      "rewards/chosen": -4.368352890014648,
      "rewards/margins": 0.0,
      "rewards/rejected": -4.368352890014648,
      "step": 30
    },
    {
      "epoch": 0.6694560669456067,
      "grad_norm": 0.6799069863533791,
      "learning_rate": 1.4248369943086995e-07,
      "logits/chosen": 1.524221420288086,
      "logits/rejected": 1.524221420288086,
      "logps/chosen": -909.3970947265625,
      "logps/pi_response": -909.3892822265625,
      "logps/ref_response": -150.93089294433594,
      "logps/rejected": -909.3970947265625,
      "loss": 0.6931,
      "rewards/accuracies": 0.0,
      "rewards/chosen": -7.585102081298828,
      "rewards/margins": 0.0,
      "rewards/rejected": -7.585102081298828,
      "step": 40
    },
    {
      "epoch": 0.8368200836820083,
      "grad_norm": 0.7614228828029587,
      "learning_rate": 3.473909705816111e-08,
      "logits/chosen": 3.145961284637451,
      "logits/rejected": 3.145961284637451,
      "logps/chosen": -1317.4720458984375,
      "logps/pi_response": -1317.4351806640625,
      "logps/ref_response": -151.639404296875,
      "logps/rejected": -1317.4720458984375,
      "loss": 0.6931,
      "rewards/accuracies": 0.0,
      "rewards/chosen": -11.658597946166992,
      "rewards/margins": 0.0,
      "rewards/rejected": -11.658597946166992,
      "step": 50
    },
    {
      "epoch": 0.9874476987447699,
      "step": 59,
      "total_flos": 0.0,
      "train_loss": 0.69314208980334,
      "train_runtime": 1040.491,
      "train_samples_per_second": 14.688,
      "train_steps_per_second": 0.057
    }
  ],
  "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
}