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": 5.776347249629427,
      "learning_rate": 8.333333333333333e-08,
      "logits/chosen": -2.832916021347046,
      "logits/rejected": -2.8954272270202637,
      "logps/chosen": -112.55461120605469,
      "logps/pi_response": -112.97522735595703,
      "logps/ref_response": -112.97522735595703,
      "logps/rejected": -126.53972625732422,
      "loss": 0.6925,
      "rewards/accuracies": 0.0,
      "rewards/chosen": 0.0,
      "rewards/margins": 0.0,
      "rewards/rejected": 0.0,
      "step": 1
    },
    {
      "epoch": 0.16736401673640167,
      "grad_norm": 5.912336712519558,
      "learning_rate": 4.930057285201027e-07,
      "logits/chosen": -2.7919647693634033,
      "logits/rejected": -2.7902114391326904,
      "logps/chosen": -150.03506469726562,
      "logps/pi_response": -152.3460693359375,
      "logps/ref_response": -152.45172119140625,
      "logps/rejected": -154.693115234375,
      "loss": 0.6924,
      "rewards/accuracies": 0.4305555522441864,
      "rewards/chosen": 0.0003976296284236014,
      "rewards/margins": -0.0001906168181449175,
      "rewards/rejected": 0.000588246330153197,
      "step": 10
    },
    {
      "epoch": 0.33472803347280333,
      "grad_norm": 6.02225171380915,
      "learning_rate": 4.187457503795526e-07,
      "logits/chosen": -2.822767972946167,
      "logits/rejected": -2.808192729949951,
      "logps/chosen": -147.93670654296875,
      "logps/pi_response": -148.9482421875,
      "logps/ref_response": -147.01174926757812,
      "logps/rejected": -150.41445922851562,
      "loss": 0.6928,
      "rewards/accuracies": 0.44999998807907104,
      "rewards/chosen": -0.022129487246274948,
      "rewards/margins": -0.0039032220374792814,
      "rewards/rejected": -0.0182262621819973,
      "step": 20
    },
    {
      "epoch": 0.502092050209205,
      "grad_norm": 6.134054284556473,
      "learning_rate": 2.8691164100062034e-07,
      "logits/chosen": -2.8247873783111572,
      "logits/rejected": -2.8113598823547363,
      "logps/chosen": -166.9477081298828,
      "logps/pi_response": -164.89968872070312,
      "logps/ref_response": -149.63009643554688,
      "logps/rejected": -166.7788848876953,
      "loss": 0.692,
      "rewards/accuracies": 0.48750001192092896,
      "rewards/chosen": -0.15240411460399628,
      "rewards/margins": -0.0001383073686156422,
      "rewards/rejected": -0.15226581692695618,
      "step": 30
    },
    {
      "epoch": 0.6694560669456067,
      "grad_norm": 6.3666889370862085,
      "learning_rate": 1.4248369943086995e-07,
      "logits/chosen": -2.7281992435455322,
      "logits/rejected": -2.721857786178589,
      "logps/chosen": -158.48544311523438,
      "logps/pi_response": -156.80209350585938,
      "logps/ref_response": -143.953857421875,
      "logps/rejected": -156.04202270507812,
      "loss": 0.6908,
      "rewards/accuracies": 0.543749988079071,
      "rewards/chosen": -0.12262190878391266,
      "rewards/margins": 0.013718548230826855,
      "rewards/rejected": -0.1363404393196106,
      "step": 40
    },
    {
      "epoch": 0.8368200836820083,
      "grad_norm": 6.157460299989396,
      "learning_rate": 3.473909705816111e-08,
      "logits/chosen": -2.74588942527771,
      "logits/rejected": -2.745689868927002,
      "logps/chosen": -152.3079833984375,
      "logps/pi_response": -150.0457000732422,
      "logps/ref_response": -140.59339904785156,
      "logps/rejected": -147.81971740722656,
      "loss": 0.6908,
      "rewards/accuracies": 0.4937500059604645,
      "rewards/chosen": -0.09621705859899521,
      "rewards/margins": 0.0019109401619061828,
      "rewards/rejected": -0.09812799841165543,
      "step": 50
    },
    {
      "epoch": 0.9874476987447699,
      "step": 59,
      "total_flos": 0.0,
      "train_loss": 0.6910505557464341,
      "train_runtime": 1298.8389,
      "train_samples_per_second": 11.767,
      "train_steps_per_second": 0.045
    }
  ],
  "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
}