trainer_state.json

{
  "best_metric": 1.6737581491470337,
  "best_model_checkpoint": "saves/Llama-3.1-8B-Instruct/lora/saa-100/checkpoint-50",
  "epoch": 8.88888888888889,
  "eval_steps": 50,
  "global_step": 50,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 1.7777777777777777,
      "grad_norm": 6.750175952911377,
      "learning_rate": 4.849231551964771e-06,
      "logits/chosen": -0.4182683825492859,
      "logits/rejected": -0.48303142189979553,
      "logps/chosen": -1.748952865600586,
      "logps/rejected": -2.1177902221679688,
      "loss": 1.8042,
      "odds_ratio_loss": 15.928056716918945,
      "rewards/accuracies": 0.762499988079071,
      "rewards/chosen": -0.1748953014612198,
      "rewards/margins": 0.03688373416662216,
      "rewards/rejected": -0.21177902817726135,
      "sft_loss": 0.21138736605644226,
      "step": 10
    },
    {
      "epoch": 3.5555555555555554,
      "grad_norm": 7.712320804595947,
      "learning_rate": 3.7500000000000005e-06,
      "logits/chosen": -0.4254830479621887,
      "logits/rejected": -0.4909093379974365,
      "logps/chosen": -1.695302963256836,
      "logps/rejected": -2.1161155700683594,
      "loss": 1.7489,
      "odds_ratio_loss": 15.380607604980469,
      "rewards/accuracies": 0.762499988079071,
      "rewards/chosen": -0.16953031718730927,
      "rewards/margins": 0.04208124801516533,
      "rewards/rejected": -0.2116115540266037,
      "sft_loss": 0.2108411341905594,
      "step": 20
    },
    {
      "epoch": 5.333333333333333,
      "grad_norm": 5.558438301086426,
      "learning_rate": 2.0658795558326745e-06,
      "logits/chosen": -0.41495656967163086,
      "logits/rejected": -0.4725929796695709,
      "logps/chosen": -1.4818356037139893,
      "logps/rejected": -1.881967306137085,
      "loss": 1.535,
      "odds_ratio_loss": 13.557626724243164,
      "rewards/accuracies": 0.7562500238418579,
      "rewards/chosen": -0.14818355441093445,
      "rewards/margins": 0.04001319035887718,
      "rewards/rejected": -0.18819674849510193,
      "sft_loss": 0.17924994230270386,
      "step": 30
    },
    {
      "epoch": 7.111111111111111,
      "grad_norm": 7.228485584259033,
      "learning_rate": 5.848888922025553e-07,
      "logits/chosen": -0.38793984055519104,
      "logits/rejected": -0.4562312662601471,
      "logps/chosen": -1.3760316371917725,
      "logps/rejected": -1.788252830505371,
      "loss": 1.4285,
      "odds_ratio_loss": 12.651687622070312,
      "rewards/accuracies": 0.768750011920929,
      "rewards/chosen": -0.13760316371917725,
      "rewards/margins": 0.041222114115953445,
      "rewards/rejected": -0.1788252890110016,
      "sft_loss": 0.1633470356464386,
      "step": 40
    },
    {
      "epoch": 8.88888888888889,
      "grad_norm": 5.6137375831604,
      "learning_rate": 0.0,
      "logits/chosen": -0.40072911977767944,
      "logits/rejected": -0.4698418974876404,
      "logps/chosen": -1.335200548171997,
      "logps/rejected": -1.7654918432235718,
      "loss": 1.3868,
      "odds_ratio_loss": 12.27790641784668,
      "rewards/accuracies": 0.7749999761581421,
      "rewards/chosen": -0.13352005183696747,
      "rewards/margins": 0.04302913695573807,
      "rewards/rejected": -0.17654919624328613,
      "sft_loss": 0.15898875892162323,
      "step": 50
    },
    {
      "epoch": 8.88888888888889,
      "eval_logits/chosen": -0.39696812629699707,
      "eval_logits/rejected": -0.5027680397033691,
      "eval_logps/chosen": -1.6218750476837158,
      "eval_logps/rejected": -2.2122323513031006,
      "eval_loss": 1.6737581491470337,
      "eval_odds_ratio_loss": 14.46813678741455,
      "eval_rewards/accuracies": 0.800000011920929,
      "eval_rewards/chosen": -0.16218750178813934,
      "eval_rewards/margins": 0.05903574824333191,
      "eval_rewards/rejected": -0.22122326493263245,
      "eval_runtime": 0.385,
      "eval_samples_per_second": 25.974,
      "eval_sft_loss": 0.22694425284862518,
      "eval_steps_per_second": 12.987,
      "step": 50
    },
    {
      "epoch": 8.88888888888889,
      "step": 50,
      "total_flos": 9214586883145728.0,
      "train_loss": 1.5806806755065919,
      "train_runtime": 281.377,
      "train_samples_per_second": 3.199,
      "train_steps_per_second": 0.178
    }
  ],
  "logging_steps": 10,
  "max_steps": 50,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 10,
  "save_steps": 50,
  "stateful_callbacks": {
    "TrainerControl": {
      "args": {
        "should_epoch_stop": false,
        "should_evaluate": false,
        "should_log": false,
        "should_save": true,
        "should_training_stop": true
      },
      "attributes": {}
    }
  },
  "total_flos": 9214586883145728.0,
  "train_batch_size": 2,
  "trial_name": null,
  "trial_params": null
}