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": 6.740061107347406,
      "learning_rate": 1.6666666666666664e-08,
      "logits/chosen": -2.7879996299743652,
      "logits/rejected": -2.789036750793457,
      "logps/chosen": -155.33309936523438,
      "logps/pi_response": -163.74407958984375,
      "logps/ref_response": -163.74407958984375,
      "logps/rejected": -160.54603576660156,
      "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": 7.101331538874614,
      "learning_rate": 9.860114570402053e-08,
      "logits/chosen": -2.732710599899292,
      "logits/rejected": -2.7246806621551514,
      "logps/chosen": -148.40945434570312,
      "logps/pi_response": -149.57220458984375,
      "logps/ref_response": -149.60110473632812,
      "logps/rejected": -148.08311462402344,
      "loss": 0.6933,
      "rewards/accuracies": 0.4513888955116272,
      "rewards/chosen": 0.00027455115923658013,
      "rewards/margins": -0.0001711228978820145,
      "rewards/rejected": 0.0004456740280147642,
      "step": 10
    },
    {
      "epoch": 0.33472803347280333,
      "grad_norm": 7.337554336815785,
      "learning_rate": 8.374915007591053e-08,
      "logits/chosen": -2.6503920555114746,
      "logits/rejected": -2.64430570602417,
      "logps/chosen": -160.72381591796875,
      "logps/pi_response": -162.48867797851562,
      "logps/ref_response": -162.63189697265625,
      "logps/rejected": -161.467041015625,
      "loss": 0.6935,
      "rewards/accuracies": 0.48750001192092896,
      "rewards/chosen": 0.0015181514900177717,
      "rewards/margins": -0.00011065408762078732,
      "rewards/rejected": 0.0016288056503981352,
      "step": 20
    },
    {
      "epoch": 0.502092050209205,
      "grad_norm": 6.797623155516534,
      "learning_rate": 5.738232820012406e-08,
      "logits/chosen": -2.6548681259155273,
      "logits/rejected": -2.6514317989349365,
      "logps/chosen": -148.1983642578125,
      "logps/pi_response": -147.19541931152344,
      "logps/ref_response": -147.06637573242188,
      "logps/rejected": -147.02438354492188,
      "loss": 0.6933,
      "rewards/accuracies": 0.4781250059604645,
      "rewards/chosen": -0.0014096560189500451,
      "rewards/margins": -0.000304248504107818,
      "rewards/rejected": -0.0011054074857383966,
      "step": 30
    },
    {
      "epoch": 0.6694560669456067,
      "grad_norm": 6.526352701379774,
      "learning_rate": 2.8496739886173992e-08,
      "logits/chosen": -2.7033307552337646,
      "logits/rejected": -2.697089433670044,
      "logps/chosen": -158.65147399902344,
      "logps/pi_response": -156.37876892089844,
      "logps/ref_response": -156.102783203125,
      "logps/rejected": -155.2120361328125,
      "loss": 0.6932,
      "rewards/accuracies": 0.4749999940395355,
      "rewards/chosen": -0.002376242307946086,
      "rewards/margins": 2.0093168132007122e-05,
      "rewards/rejected": -0.0023963353596627712,
      "step": 40
    },
    {
      "epoch": 0.8368200836820083,
      "grad_norm": 6.8095971653212,
      "learning_rate": 6.947819411632222e-09,
      "logits/chosen": -2.687405586242676,
      "logits/rejected": -2.6960949897766113,
      "logps/chosen": -156.6856689453125,
      "logps/pi_response": -156.02719116210938,
      "logps/ref_response": -155.70803833007812,
      "logps/rejected": -160.33645629882812,
      "loss": 0.6932,
      "rewards/accuracies": 0.4781250059604645,
      "rewards/chosen": -0.002912611234933138,
      "rewards/margins": 0.0002370561269344762,
      "rewards/rejected": -0.0031496677547693253,
      "step": 50
    },
    {
      "epoch": 0.9874476987447699,
      "step": 59,
      "total_flos": 0.0,
      "train_loss": 0.6932335586871131,
      "train_runtime": 2785.4094,
      "train_samples_per_second": 5.487,
      "train_steps_per_second": 0.021
    }
  ],
  "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
}