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": 11.88069732420037,
      "learning_rate": 8.333333333333333e-08,
      "logits/chosen": -2.8434743881225586,
      "logits/rejected": -2.769583225250244,
      "logps/chosen": -178.4530487060547,
      "logps/pi_response": -97.4625244140625,
      "logps/ref_response": -97.4625244140625,
      "logps/rejected": -233.52891540527344,
      "loss": 0.6956,
      "rewards/accuracies": 0.0,
      "rewards/chosen": 0.0,
      "rewards/margins": 0.0,
      "rewards/rejected": 0.0,
      "step": 1
    },
    {
      "epoch": 0.16736401673640167,
      "grad_norm": 11.505588607871543,
      "learning_rate": 4.930057285201027e-07,
      "logits/chosen": -2.7551352977752686,
      "logits/rejected": -2.7301125526428223,
      "logps/chosen": -225.8229522705078,
      "logps/pi_response": -120.52629852294922,
      "logps/ref_response": -119.74172973632812,
      "logps/rejected": -265.8622131347656,
      "loss": 0.6875,
      "rewards/accuracies": 0.5972222089767456,
      "rewards/chosen": -0.03128578141331673,
      "rewards/margins": 0.02152625285089016,
      "rewards/rejected": -0.05281204357743263,
      "step": 10
    },
    {
      "epoch": 0.33472803347280333,
      "grad_norm": 25.557753221670836,
      "learning_rate": 4.187457503795526e-07,
      "logits/chosen": -2.8003687858581543,
      "logits/rejected": -2.7502455711364746,
      "logps/chosen": -259.4373474121094,
      "logps/pi_response": -115.7847900390625,
      "logps/ref_response": -118.667724609375,
      "logps/rejected": -311.7029113769531,
      "loss": 0.6364,
      "rewards/accuracies": 0.643750011920929,
      "rewards/chosen": -0.24674002826213837,
      "rewards/margins": 0.29084205627441406,
      "rewards/rejected": -0.5375820994377136,
      "step": 20
    },
    {
      "epoch": 0.502092050209205,
      "grad_norm": 15.538988265977032,
      "learning_rate": 2.8691164100062034e-07,
      "logits/chosen": -2.782900094985962,
      "logits/rejected": -2.7508139610290527,
      "logps/chosen": -289.4466247558594,
      "logps/pi_response": -149.1678466796875,
      "logps/ref_response": -124.9267807006836,
      "logps/rejected": -358.2657470703125,
      "loss": 0.5877,
      "rewards/accuracies": 0.7124999761581421,
      "rewards/chosen": -0.4117729663848877,
      "rewards/margins": 0.5106114745140076,
      "rewards/rejected": -0.9223844408988953,
      "step": 30
    },
    {
      "epoch": 0.6694560669456067,
      "grad_norm": 19.459335671099513,
      "learning_rate": 1.4248369943086995e-07,
      "logits/chosen": -2.680762767791748,
      "logits/rejected": -2.6408679485321045,
      "logps/chosen": -269.59991455078125,
      "logps/pi_response": -149.72483825683594,
      "logps/ref_response": -115.91612243652344,
      "logps/rejected": -360.96343994140625,
      "loss": 0.5607,
      "rewards/accuracies": 0.706250011920929,
      "rewards/chosen": -0.552818238735199,
      "rewards/margins": 0.5624674558639526,
      "rewards/rejected": -1.1152856349945068,
      "step": 40
    },
    {
      "epoch": 0.8368200836820083,
      "grad_norm": 21.120309284129263,
      "learning_rate": 3.473909705816111e-08,
      "logits/chosen": -2.7053980827331543,
      "logits/rejected": -2.6527323722839355,
      "logps/chosen": -259.5418701171875,
      "logps/pi_response": -140.0457305908203,
      "logps/ref_response": -106.76103210449219,
      "logps/rejected": -333.9945373535156,
      "loss": 0.5657,
      "rewards/accuracies": 0.706250011920929,
      "rewards/chosen": -0.5184835195541382,
      "rewards/margins": 0.5243555307388306,
      "rewards/rejected": -1.0428390502929688,
      "step": 50
    },
    {
      "epoch": 0.9874476987447699,
      "step": 59,
      "total_flos": 0.0,
      "train_loss": 0.6013881675267624,
      "train_runtime": 1334.1164,
      "train_samples_per_second": 11.456,
      "train_steps_per_second": 0.044
    }
  ],
  "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
}