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": 5.772511137926767,
      "learning_rate": 8.333333333333333e-08,
      "logits/chosen": -2.801664352416992,
      "logits/rejected": -2.8308780193328857,
      "logps/chosen": -137.20802307128906,
      "logps/pi_response": -133.3135223388672,
      "logps/ref_response": -133.3135223388672,
      "logps/rejected": -141.6568145751953,
      "loss": 0.6932,
      "rewards/accuracies": 0.0,
      "rewards/chosen": 0.0,
      "rewards/margins": 0.0,
      "rewards/rejected": 0.0,
      "step": 1
    },
    {
      "epoch": 0.16736401673640167,
      "grad_norm": 5.866571559196428,
      "learning_rate": 4.930057285201027e-07,
      "logits/chosen": -2.7963476181030273,
      "logits/rejected": -2.7924842834472656,
      "logps/chosen": -155.5491943359375,
      "logps/pi_response": -153.5810089111328,
      "logps/ref_response": -153.69000244140625,
      "logps/rejected": -155.2954559326172,
      "loss": 0.6932,
      "rewards/accuracies": 0.4340277910232544,
      "rewards/chosen": 0.0005564650637097657,
      "rewards/margins": -0.00014017036301083863,
      "rewards/rejected": 0.0006966353976167738,
      "step": 10
    },
    {
      "epoch": 0.33472803347280333,
      "grad_norm": 5.991267980339142,
      "learning_rate": 4.187457503795526e-07,
      "logits/chosen": -2.828132152557373,
      "logits/rejected": -2.8157670497894287,
      "logps/chosen": -149.9945526123047,
      "logps/pi_response": -149.78407287597656,
      "logps/ref_response": -147.89337158203125,
      "logps/rejected": -152.16629028320312,
      "loss": 0.6936,
      "rewards/accuracies": 0.4906249940395355,
      "rewards/chosen": -0.019678335636854172,
      "rewards/margins": -8.539492409909144e-05,
      "rewards/rejected": -0.01959294266998768,
      "step": 20
    },
    {
      "epoch": 0.502092050209205,
      "grad_norm": 6.102937202185575,
      "learning_rate": 2.8691164100062034e-07,
      "logits/chosen": -2.784358263015747,
      "logits/rejected": -2.7720556259155273,
      "logps/chosen": -156.15835571289062,
      "logps/pi_response": -155.635009765625,
      "logps/ref_response": -141.50277709960938,
      "logps/rejected": -156.25270080566406,
      "loss": 0.6926,
      "rewards/accuracies": 0.503125011920929,
      "rewards/chosen": -0.13776005804538727,
      "rewards/margins": 0.00394659535959363,
      "rewards/rejected": -0.1417066603899002,
      "step": 30
    },
    {
      "epoch": 0.6694560669456067,
      "grad_norm": 6.597531175861638,
      "learning_rate": 1.4248369943086995e-07,
      "logits/chosen": -2.7550230026245117,
      "logits/rejected": -2.753791093826294,
      "logps/chosen": -157.1768341064453,
      "logps/pi_response": -156.76754760742188,
      "logps/ref_response": -143.01797485351562,
      "logps/rejected": -155.65341186523438,
      "loss": 0.6917,
      "rewards/accuracies": 0.515625,
      "rewards/chosen": -0.13175955414772034,
      "rewards/margins": 0.008598078042268753,
      "rewards/rejected": -0.1403576284646988,
      "step": 40
    },
    {
      "epoch": 0.8368200836820083,
      "grad_norm": 6.081555901091694,
      "learning_rate": 3.473909705816111e-08,
      "logits/chosen": -2.7671895027160645,
      "logits/rejected": -2.767414093017578,
      "logps/chosen": -149.20730590820312,
      "logps/pi_response": -149.70562744140625,
      "logps/ref_response": -138.1813201904297,
      "logps/rejected": -146.49392700195312,
      "loss": 0.6918,
      "rewards/accuracies": 0.4937500059604645,
      "rewards/chosen": -0.11548925936222076,
      "rewards/margins": 0.0021973345428705215,
      "rewards/rejected": -0.11768659204244614,
      "step": 50
    },
    {
      "epoch": 0.9874476987447699,
      "step": 59,
      "total_flos": 0.0,
      "train_loss": 0.6921672336125778,
      "train_runtime": 2751.1622,
      "train_samples_per_second": 5.555,
      "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
}