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": 5.810358768483852,
      "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": 0.6837,
      "rewards/accuracies": 0.0,
      "rewards/chosen": 0.0,
      "rewards/margins": 0.0,
      "rewards/rejected": 0.0,
      "step": 1
    },
    {
      "epoch": 0.16736401673640167,
      "grad_norm": 7.406448882690957,
      "learning_rate": 4.930057285201027e-07,
      "logits/chosen": -2.7853667736053467,
      "logits/rejected": -2.78360652923584,
      "logps/chosen": -151.95591735839844,
      "logps/pi_response": -154.27581787109375,
      "logps/ref_response": -152.45172119140625,
      "logps/rejected": -156.6210174560547,
      "loss": 0.6828,
      "rewards/accuracies": 0.3958333432674408,
      "rewards/chosen": -0.018810953944921494,
      "rewards/margins": -0.00012028714263578877,
      "rewards/rejected": -0.018690669909119606,
      "step": 10
    },
    {
      "epoch": 0.33472803347280333,
      "grad_norm": 5.646110166139303,
      "learning_rate": 4.187457503795526e-07,
      "logits/chosen": -2.7280337810516357,
      "logits/rejected": -2.71278715133667,
      "logps/chosen": -180.63790893554688,
      "logps/pi_response": -181.5164337158203,
      "logps/ref_response": -147.01174926757812,
      "logps/rejected": -182.7357635498047,
      "loss": 0.6834,
      "rewards/accuracies": 0.4749999940395355,
      "rewards/chosen": -0.3491414785385132,
      "rewards/margins": -0.007702186703681946,
      "rewards/rejected": -0.34143924713134766,
      "step": 20
    },
    {
      "epoch": 0.502092050209205,
      "grad_norm": 6.304075320741244,
      "learning_rate": 2.8691164100062034e-07,
      "logits/chosen": -2.5796682834625244,
      "logits/rejected": -2.568051338195801,
      "logps/chosen": -230.05471801757812,
      "logps/pi_response": -227.27792358398438,
      "logps/ref_response": -149.63009643554688,
      "logps/rejected": -229.08181762695312,
      "loss": 0.6818,
      "rewards/accuracies": 0.44999998807907104,
      "rewards/chosen": -0.783474326133728,
      "rewards/margins": -0.008179178461432457,
      "rewards/rejected": -0.7752951383590698,
      "step": 30
    },
    {
      "epoch": 0.6694560669456067,
      "grad_norm": 8.014543776641343,
      "learning_rate": 1.4248369943086995e-07,
      "logits/chosen": -2.330618381500244,
      "logits/rejected": -2.3253395557403564,
      "logps/chosen": -284.21649169921875,
      "logps/pi_response": -280.53057861328125,
      "logps/ref_response": -143.953857421875,
      "logps/rejected": -282.0881652832031,
      "loss": 0.6756,
      "rewards/accuracies": 0.512499988079071,
      "rewards/chosen": -1.3799324035644531,
      "rewards/margins": 0.016869211569428444,
      "rewards/rejected": -1.3968017101287842,
      "step": 40
    },
    {
      "epoch": 0.8368200836820083,
      "grad_norm": 8.890401678686317,
      "learning_rate": 3.473909705816111e-08,
      "logits/chosen": -2.233668088912964,
      "logits/rejected": -2.23595929145813,
      "logps/chosen": -339.37506103515625,
      "logps/pi_response": -335.8888854980469,
      "logps/ref_response": -140.59339904785156,
      "logps/rejected": -329.88592529296875,
      "loss": 0.6822,
      "rewards/accuracies": 0.48750001192092896,
      "rewards/chosen": -1.9668878316879272,
      "rewards/margins": -0.048098064959049225,
      "rewards/rejected": -1.9187901020050049,
      "step": 50
    },
    {
      "epoch": 0.9874476987447699,
      "step": 59,
      "total_flos": 0.0,
      "train_loss": 0.6825068279848261,
      "train_runtime": 1304.8348,
      "train_samples_per_second": 11.713,
      "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
}