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": 53.94166685078156,
      "learning_rate": 8.333333333333333e-08,
      "logits/chosen": -2.6333391666412354,
      "logits/rejected": -2.579299211502075,
      "logps/chosen": -217.46055603027344,
      "logps/pi_response": -144.18421936035156,
      "logps/ref_response": -144.18421936035156,
      "logps/rejected": -451.1424560546875,
      "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": 36.91625770695295,
      "learning_rate": 4.930057285201027e-07,
      "logits/chosen": -2.593613862991333,
      "logits/rejected": -2.541897773742676,
      "logps/chosen": -269.9593200683594,
      "logps/pi_response": -130.84666442871094,
      "logps/ref_response": -122.69916534423828,
      "logps/rejected": -420.3211669921875,
      "loss": 0.6626,
      "rewards/accuracies": 0.5763888955116272,
      "rewards/chosen": -0.2059764266014099,
      "rewards/margins": 0.11146806180477142,
      "rewards/rejected": -0.31744447350502014,
      "step": 10
    },
    {
      "epoch": 0.33472803347280333,
      "grad_norm": 41.02564046309243,
      "learning_rate": 4.187457503795526e-07,
      "logits/chosen": -2.4761414527893066,
      "logits/rejected": -2.398340940475464,
      "logps/chosen": -311.044677734375,
      "logps/pi_response": -148.96127319335938,
      "logps/ref_response": -130.3125457763672,
      "logps/rejected": -574.2133178710938,
      "loss": 0.5752,
      "rewards/accuracies": 0.7281249761581421,
      "rewards/chosen": -0.7746369242668152,
      "rewards/margins": 1.01192045211792,
      "rewards/rejected": -1.7865574359893799,
      "step": 20
    },
    {
      "epoch": 0.502092050209205,
      "grad_norm": 35.01071098541561,
      "learning_rate": 2.8691164100062034e-07,
      "logits/chosen": -2.3740668296813965,
      "logits/rejected": -2.304201602935791,
      "logps/chosen": -355.04644775390625,
      "logps/pi_response": -148.07662963867188,
      "logps/ref_response": -125.0134506225586,
      "logps/rejected": -537.8699340820312,
      "loss": 0.5296,
      "rewards/accuracies": 0.721875011920929,
      "rewards/chosen": -0.8583258390426636,
      "rewards/margins": 0.7140148282051086,
      "rewards/rejected": -1.572340726852417,
      "step": 30
    },
    {
      "epoch": 0.6694560669456067,
      "grad_norm": 29.40836483007019,
      "learning_rate": 1.4248369943086995e-07,
      "logits/chosen": -2.3838210105895996,
      "logits/rejected": -2.2967305183410645,
      "logps/chosen": -310.45281982421875,
      "logps/pi_response": -141.27828979492188,
      "logps/ref_response": -130.2584991455078,
      "logps/rejected": -555.8219604492188,
      "loss": 0.5076,
      "rewards/accuracies": 0.78125,
      "rewards/chosen": -0.5878928899765015,
      "rewards/margins": 0.7744524478912354,
      "rewards/rejected": -1.3623450994491577,
      "step": 40
    },
    {
      "epoch": 0.8368200836820083,
      "grad_norm": 28.867819137396587,
      "learning_rate": 3.473909705816111e-08,
      "logits/chosen": -2.3730177879333496,
      "logits/rejected": -2.3055779933929443,
      "logps/chosen": -307.48468017578125,
      "logps/pi_response": -135.36683654785156,
      "logps/ref_response": -127.3092269897461,
      "logps/rejected": -545.8276977539062,
      "loss": 0.5015,
      "rewards/accuracies": 0.762499988079071,
      "rewards/chosen": -0.5906313061714172,
      "rewards/margins": 0.8331750631332397,
      "rewards/rejected": -1.4238064289093018,
      "step": 50
    },
    {
      "epoch": 0.9874476987447699,
      "step": 59,
      "total_flos": 0.0,
      "train_loss": 0.5476259296223268,
      "train_runtime": 2608.0789,
      "train_samples_per_second": 5.86,
      "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
}