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": 77.33576785231257,
      "learning_rate": 1.6666666666666664e-08,
      "logits/chosen": -2.34324312210083,
      "logits/rejected": -2.300614356994629,
      "logps/chosen": -266.8575439453125,
      "logps/pi_response": -149.64938354492188,
      "logps/ref_response": -149.64938354492188,
      "logps/rejected": -583.212890625,
      "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": 79.71187437832909,
      "learning_rate": 9.860114570402053e-08,
      "logits/chosen": -2.344362258911133,
      "logits/rejected": -2.2582879066467285,
      "logps/chosen": -289.9632263183594,
      "logps/pi_response": -132.39773559570312,
      "logps/ref_response": -131.7936248779297,
      "logps/rejected": -560.7911376953125,
      "loss": 0.6858,
      "rewards/accuracies": 0.5763888955116272,
      "rewards/chosen": -0.01403072103857994,
      "rewards/margins": 0.012026474811136723,
      "rewards/rejected": -0.026057196781039238,
      "step": 10
    },
    {
      "epoch": 0.33472803347280333,
      "grad_norm": 56.29950708719972,
      "learning_rate": 8.374915007591053e-08,
      "logits/chosen": -2.296107769012451,
      "logits/rejected": -2.2206854820251465,
      "logps/chosen": -296.99676513671875,
      "logps/pi_response": -146.00672912597656,
      "logps/ref_response": -139.09341430664062,
      "logps/rejected": -597.2777709960938,
      "loss": 0.6189,
      "rewards/accuracies": 0.706250011920929,
      "rewards/chosen": -0.2067621946334839,
      "rewards/margins": 0.18077115714550018,
      "rewards/rejected": -0.3875333368778229,
      "step": 20
    },
    {
      "epoch": 0.502092050209205,
      "grad_norm": 33.48276555701894,
      "learning_rate": 5.738232820012406e-08,
      "logits/chosen": -2.2276177406311035,
      "logits/rejected": -2.1680381298065186,
      "logps/chosen": -329.71270751953125,
      "logps/pi_response": -141.6411895751953,
      "logps/ref_response": -133.272216796875,
      "logps/rejected": -643.0191650390625,
      "loss": 0.5628,
      "rewards/accuracies": 0.737500011920929,
      "rewards/chosen": -0.3608744144439697,
      "rewards/margins": 0.4366278648376465,
      "rewards/rejected": -0.797502338886261,
      "step": 30
    },
    {
      "epoch": 0.6694560669456067,
      "grad_norm": 35.81791552796474,
      "learning_rate": 2.8496739886173992e-08,
      "logits/chosen": -2.2278857231140137,
      "logits/rejected": -2.1367008686065674,
      "logps/chosen": -357.0513610839844,
      "logps/pi_response": -151.4658660888672,
      "logps/ref_response": -141.07730102539062,
      "logps/rejected": -702.927978515625,
      "loss": 0.5193,
      "rewards/accuracies": 0.7281249761581421,
      "rewards/chosen": -0.5399668216705322,
      "rewards/margins": 0.7153794765472412,
      "rewards/rejected": -1.2553462982177734,
      "step": 40
    },
    {
      "epoch": 0.8368200836820083,
      "grad_norm": 83.14214128894021,
      "learning_rate": 6.947819411632222e-09,
      "logits/chosen": -2.1951634883880615,
      "logits/rejected": -2.135693311691284,
      "logps/chosen": -356.71759033203125,
      "logps/pi_response": -150.52035522460938,
      "logps/ref_response": -138.71400451660156,
      "logps/rejected": -741.8605346679688,
      "loss": 0.5157,
      "rewards/accuracies": 0.7437499761581421,
      "rewards/chosen": -0.5678369402885437,
      "rewards/margins": 0.8946306109428406,
      "rewards/rejected": -1.4624677896499634,
      "step": 50
    },
    {
      "epoch": 0.9874476987447699,
      "step": 59,
      "total_flos": 0.0,
      "train_loss": 0.5653046753446934,
      "train_runtime": 2582.5246,
      "train_samples_per_second": 5.918,
      "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
}