trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 0.9905956112852664,
  "eval_steps": 500,
  "global_step": 79,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.01,
      "learning_rate": 6.25e-08,
      "logits/chosen": -2.9087584018707275,
      "logits/rejected": -2.8338208198547363,
      "logps/chosen": -352.96319580078125,
      "logps/pi_response": -77.43819427490234,
      "logps/ref_response": -77.43819427490234,
      "logps/rejected": -170.26690673828125,
      "loss": 0.6931,
      "rewards/accuracies": 0.0,
      "rewards/chosen": 0.0,
      "rewards/margins": 0.0,
      "rewards/rejected": 0.0,
      "step": 1
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.990217055187362e-07,
      "logits/chosen": -2.787048101425171,
      "logits/rejected": -2.764052391052246,
      "logps/chosen": -237.8766326904297,
      "logps/pi_response": -73.74913024902344,
      "logps/ref_response": -73.52912139892578,
      "logps/rejected": -170.56431579589844,
      "loss": 0.6907,
      "rewards/accuracies": 0.5451388955116272,
      "rewards/chosen": 0.005484491586685181,
      "rewards/margins": 0.004212961066514254,
      "rewards/rejected": 0.0012715309858322144,
      "step": 10
    },
    {
      "epoch": 0.25,
      "learning_rate": 4.655786431300069e-07,
      "logits/chosen": -2.7454662322998047,
      "logits/rejected": -2.6918036937713623,
      "logps/chosen": -248.79116821289062,
      "logps/pi_response": -79.92262268066406,
      "logps/ref_response": -71.4201431274414,
      "logps/rejected": -174.75128173828125,
      "loss": 0.6604,
      "rewards/accuracies": 0.715624988079071,
      "rewards/chosen": 0.04207359999418259,
      "rewards/margins": 0.0811639204621315,
      "rewards/rejected": -0.039090320467948914,
      "step": 20
    },
    {
      "epoch": 0.38,
      "learning_rate": 3.9061232191019517e-07,
      "logits/chosen": -2.6588845252990723,
      "logits/rejected": -2.6236281394958496,
      "logps/chosen": -234.9491729736328,
      "logps/pi_response": -105.2069320678711,
      "logps/ref_response": -68.3199691772461,
      "logps/rejected": -190.48977661132812,
      "loss": 0.6266,
      "rewards/accuracies": 0.7406250238418579,
      "rewards/chosen": -0.09634245932102203,
      "rewards/margins": 0.1879693567752838,
      "rewards/rejected": -0.28431180119514465,
      "step": 30
    },
    {
      "epoch": 0.5,
      "learning_rate": 2.8856223324132555e-07,
      "logits/chosen": -2.665360450744629,
      "logits/rejected": -2.64331316947937,
      "logps/chosen": -259.9814453125,
      "logps/pi_response": -145.37118530273438,
      "logps/ref_response": -75.2870864868164,
      "logps/rejected": -230.64920043945312,
      "loss": 0.5797,
      "rewards/accuracies": 0.746874988079071,
      "rewards/chosen": -0.22867317497730255,
      "rewards/margins": 0.351193904876709,
      "rewards/rejected": -0.5798671245574951,
      "step": 40
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.7908455541642582e-07,
      "logits/chosen": -2.667282819747925,
      "logits/rejected": -2.633934497833252,
      "logps/chosen": -287.79718017578125,
      "logps/pi_response": -155.01898193359375,
      "logps/ref_response": -74.21197509765625,
      "logps/rejected": -249.22134399414062,
      "loss": 0.5499,
      "rewards/accuracies": 0.800000011920929,
      "rewards/chosen": -0.3261149525642395,
      "rewards/margins": 0.43611159920692444,
      "rewards/rejected": -0.7622265815734863,
      "step": 50
    },
    {
      "epoch": 0.75,
      "learning_rate": 8.32661172908373e-08,
      "logits/chosen": -2.6649065017700195,
      "logits/rejected": -2.6298487186431885,
      "logps/chosen": -257.4251403808594,
      "logps/pi_response": -161.20509338378906,
      "logps/ref_response": -66.34608459472656,
      "logps/rejected": -266.110107421875,
      "loss": 0.5296,
      "rewards/accuracies": 0.793749988079071,
      "rewards/chosen": -0.42365479469299316,
      "rewards/margins": 0.5627300143241882,
      "rewards/rejected": -0.9863847494125366,
      "step": 60
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.956279997278043e-08,
      "logits/chosen": -2.6613144874572754,
      "logits/rejected": -2.624793529510498,
      "logps/chosen": -307.1275329589844,
      "logps/pi_response": -187.70135498046875,
      "logps/ref_response": -77.6395034790039,
      "logps/rejected": -278.43206787109375,
      "loss": 0.5188,
      "rewards/accuracies": 0.768750011920929,
      "rewards/chosen": -0.4925355017185211,
      "rewards/margins": 0.6049482822418213,
      "rewards/rejected": -1.0974838733673096,
      "step": 70
    },
    {
      "epoch": 0.99,
      "step": 79,
      "total_flos": 0.0,
      "train_loss": 0.586772628977329,
      "train_runtime": 4664.074,
      "train_samples_per_second": 4.369,
      "train_steps_per_second": 0.017
    }
  ],
  "logging_steps": 10,
  "max_steps": 79,
  "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
}