checkpoint-2/trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "episode": 128,
  "epoch": 1.28,
  "eval_steps": 100,
  "global_step": 2,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "episode": 64,
      "epoch": 0.64,
      "eps": 0,
      "loss/policy_avg": -0.015076532028615475,
      "loss/value_avg": 2.0768136978149414,
      "lr": 3e-06,
      "objective/entropy": 13.955379486083984,
      "objective/kl": 0.8348603248596191,
      "objective/non_score_reward": -0.041743017733097076,
      "objective/rlhf_reward": -1.4829208850860596,
      "objective/scores": -1.4411778450012207,
      "policy/approxkl_avg": 0.003379022004082799,
      "policy/clipfrac_avg": 0.02070312574505806,
      "policy/entropy_avg": 0.9434927701950073,
      "step": 1,
      "val/clipfrac_avg": 0.6554687023162842,
      "val/num_eos_tokens": 0,
      "val/ratio": 0.9607726335525513,
      "val/ratio_var": 0.0007270940695889294
    },
    {
      "episode": 128,
      "epoch": 1.28,
      "eps": 0,
      "loss/policy_avg": -0.010912335477769375,
      "loss/value_avg": 0.6733213663101196,
      "lr": 1.5e-06,
      "objective/entropy": 10.704127311706543,
      "objective/kl": 0.7690310478210449,
      "objective/non_score_reward": -0.038451552391052246,
      "objective/rlhf_reward": -1.1398342847824097,
      "objective/scores": -1.1013827323913574,
      "policy/approxkl_avg": 0.0029775083530694246,
      "policy/clipfrac_avg": 0.01894531212747097,
      "policy/entropy_avg": 0.7254143357276917,
      "step": 2,
      "val/clipfrac_avg": 0.6109374761581421,
      "val/num_eos_tokens": 0,
      "val/ratio": 0.9696837663650513,
      "val/ratio_var": 0.0008283399511128664
    }
  ],
  "logging_steps": 500,
  "max_steps": 2,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 1.0,
  "save_steps": 500,
  "stateful_callbacks": {
    "TrainerControl": {
      "args": {
        "should_epoch_stop": false,
        "should_evaluate": false,
        "should_log": false,
        "should_save": true,
        "should_training_stop": true
      },
      "attributes": {}
    }
  },
  "total_flos": 0,
  "train_batch_size": null,
  "trial_name": null,
  "trial_params": null
}