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": 0.02,
      "learning_rate": 8.333333333333334e-10,
      "logits/chosen": -2.5598204135894775,
      "logits/rejected": -2.54148006439209,
      "logps/chosen": -231.18460083007812,
      "logps/pi_response": -153.6551971435547,
      "logps/ref_response": -153.6551971435547,
      "logps/rejected": -481.5467529296875,
      "loss": 0.6931,
      "rewards/accuracies": 0.0,
      "rewards/chosen": 0.0,
      "rewards/margins": 0.0,
      "rewards/rejected": 0.0,
      "step": 1
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.930057285201027e-09,
      "logits/chosen": -2.579479455947876,
      "logits/rejected": -2.5000054836273193,
      "logps/chosen": -276.4933776855469,
      "logps/pi_response": -135.43942260742188,
      "logps/ref_response": -135.50714111328125,
      "logps/rejected": -483.53948974609375,
      "loss": 0.6932,
      "rewards/accuracies": 0.4479166567325592,
      "rewards/chosen": 0.0006561667541973293,
      "rewards/margins": -0.0003241149242967367,
      "rewards/rejected": 0.0009802815038710833,
      "step": 10
    },
    {
      "epoch": 0.33,
      "learning_rate": 4.1874575037955265e-09,
      "logits/chosen": -2.5438618659973145,
      "logits/rejected": -2.500082015991211,
      "logps/chosen": -286.8203430175781,
      "logps/pi_response": -147.23228454589844,
      "logps/ref_response": -147.16122436523438,
      "logps/rejected": -460.07977294921875,
      "loss": 0.6927,
      "rewards/accuracies": 0.53125,
      "rewards/chosen": -0.0012539768358692527,
      "rewards/margins": 0.0011572205694392323,
      "rewards/rejected": -0.002411197405308485,
      "step": 20
    },
    {
      "epoch": 0.5,
      "learning_rate": 2.8691164100062035e-09,
      "logits/chosen": -2.528888702392578,
      "logits/rejected": -2.482515811920166,
      "logps/chosen": -270.5335388183594,
      "logps/pi_response": -141.44882202148438,
      "logps/ref_response": -141.3872528076172,
      "logps/rejected": -486.7137145996094,
      "loss": 0.6925,
      "rewards/accuracies": 0.581250011920929,
      "rewards/chosen": -0.00033235797309316695,
      "rewards/margins": 0.0026473526377230883,
      "rewards/rejected": -0.0029797102324664593,
      "step": 30
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.4248369943086997e-09,
      "logits/chosen": -2.550001621246338,
      "logits/rejected": -2.5071542263031006,
      "logps/chosen": -279.8554992675781,
      "logps/pi_response": -145.85305786132812,
      "logps/ref_response": -145.77267456054688,
      "logps/rejected": -492.9515075683594,
      "loss": 0.6916,
      "rewards/accuracies": 0.550000011920929,
      "rewards/chosen": -0.003366402117535472,
      "rewards/margins": 0.0020419852808117867,
      "rewards/rejected": -0.005408387165516615,
      "step": 40
    },
    {
      "epoch": 0.84,
      "learning_rate": 3.473909705816111e-10,
      "logits/chosen": -2.5637478828430176,
      "logits/rejected": -2.5069830417633057,
      "logps/chosen": -301.5505065917969,
      "logps/pi_response": -149.0933380126953,
      "logps/ref_response": -148.97946166992188,
      "logps/rejected": -506.888671875,
      "loss": 0.6909,
      "rewards/accuracies": 0.6468750238418579,
      "rewards/chosen": -0.004720722325146198,
      "rewards/margins": 0.004021670203655958,
      "rewards/rejected": -0.008742392994463444,
      "step": 50
    },
    {
      "epoch": 0.99,
      "step": 59,
      "total_flos": 0.0,
      "train_loss": 0.6919118428634385,
      "train_runtime": 3481.2266,
      "train_samples_per_second": 4.39,
      "train_steps_per_second": 0.017
    }
  ],
  "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
}