checkpoint-100/trainer_state.json

{
  "best_metric": 0.5667449831962585,
  "best_model_checkpoint": "./Mistral/13-03-24-Weni-ZeroShot-3.4.6-Mistral-7b-DPO-1.0.0_ZeroShot DPO Training a improved dataset and best hyperparameters found so far-2_max_steps-288_batch_32_2024-03-13_ppid_9/checkpoint-100",
  "epoch": 2.0408163265306123,
  "eval_steps": 100,
  "global_step": 100,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.41,
      "grad_norm": 4.933152675628662,
      "learning_rate": 1.3793103448275862e-06,
      "logits/chosen": -1.3310586214065552,
      "logits/rejected": -1.2998794317245483,
      "logps/chosen": -17.44550895690918,
      "logps/rejected": -15.287511825561523,
      "loss": 0.6887,
      "rewards/accuracies": 0.6078125238418579,
      "rewards/chosen": 0.012908421456813812,
      "rewards/margins": 0.00912781897932291,
      "rewards/rejected": 0.0037806027103215456,
      "step": 20
    },
    {
      "epoch": 0.82,
      "grad_norm": 1.840120553970337,
      "learning_rate": 1.915057915057915e-06,
      "logits/chosen": -1.342775583267212,
      "logits/rejected": -1.3136231899261475,
      "logps/chosen": -16.57306480407715,
      "logps/rejected": -15.076986312866211,
      "loss": 0.6579,
      "rewards/accuracies": 0.7734375,
      "rewards/chosen": 0.10680261999368668,
      "rewards/margins": 0.08031970262527466,
      "rewards/rejected": 0.02648291550576687,
      "step": 40
    },
    {
      "epoch": 1.22,
      "grad_norm": 2.2041518688201904,
      "learning_rate": 1.7606177606177606e-06,
      "logits/chosen": -1.3282166719436646,
      "logits/rejected": -1.300879716873169,
      "logps/chosen": -15.91553783416748,
      "logps/rejected": -15.11401653289795,
      "loss": 0.6314,
      "rewards/accuracies": 0.785937488079071,
      "rewards/chosen": 0.16994965076446533,
      "rewards/margins": 0.14337585866451263,
      "rewards/rejected": 0.026573771610856056,
      "step": 60
    },
    {
      "epoch": 1.63,
      "grad_norm": 1.4858603477478027,
      "learning_rate": 1.606177606177606e-06,
      "logits/chosen": -1.3553317785263062,
      "logits/rejected": -1.3338539600372314,
      "logps/chosen": -15.328280448913574,
      "logps/rejected": -15.158895492553711,
      "loss": 0.6066,
      "rewards/accuracies": 0.768750011920929,
      "rewards/chosen": 0.2215496301651001,
      "rewards/margins": 0.20472590625286102,
      "rewards/rejected": 0.016823694109916687,
      "step": 80
    },
    {
      "epoch": 2.04,
      "grad_norm": 1.4387667179107666,
      "learning_rate": 1.4517374517374517e-06,
      "logits/chosen": -1.3066260814666748,
      "logits/rejected": -1.282898187637329,
      "logps/chosen": -14.973272323608398,
      "logps/rejected": -15.511428833007812,
      "loss": 0.5697,
      "rewards/accuracies": 0.793749988079071,
      "rewards/chosen": 0.2805514931678772,
      "rewards/margins": 0.2973305881023407,
      "rewards/rejected": -0.016779109835624695,
      "step": 100
    },
    {
      "epoch": 2.04,
      "eval_logits/chosen": -1.3297994136810303,
      "eval_logits/rejected": -1.3079336881637573,
      "eval_logps/chosen": -14.785962104797363,
      "eval_logps/rejected": -15.702004432678223,
      "eval_loss": 0.5667449831962585,
      "eval_rewards/accuracies": 0.7651515603065491,
      "eval_rewards/chosen": 0.30174848437309265,
      "eval_rewards/margins": 0.31671077013015747,
      "eval_rewards/rejected": -0.014962326735258102,
      "eval_runtime": 81.5286,
      "eval_samples_per_second": 2.134,
      "eval_steps_per_second": 0.27,
      "step": 100
    }
  ],
  "logging_steps": 20,
  "max_steps": 288,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 6,
  "save_steps": 100,
  "total_flos": 0.0,
  "train_batch_size": 8,
  "trial_name": null,
  "trial_params": null
}