checkpoint-50/trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 0.25690430314707774,
  "eval_steps": 500,
  "global_step": 50,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.05,
      "grad_norm": 6.844920635223389,
      "learning_rate": 5e-05,
      "logits/chosen": -3.09375,
      "logits/rejected": -3.0625,
      "logps/chosen": -258.0,
      "logps/rejected": -322.0,
      "loss": 1.0697,
      "max_memory_allocated (GB)": 94.25,
      "memory_allocated (GB)": 51.7,
      "rewards/accuracies": 0.30078125,
      "rewards/chosen": 0.67578125,
      "rewards/margins": -0.4140625,
      "rewards/rejected": 1.0859375,
      "step": 10,
      "total_memory_available (GB)": 94.62
    },
    {
      "epoch": 0.1,
      "grad_norm": 4.300484657287598,
      "learning_rate": 0.0001,
      "logits/chosen": -3.046875,
      "logits/rejected": -3.03125,
      "logps/chosen": -286.0,
      "logps/rejected": -336.0,
      "loss": 0.5736,
      "max_memory_allocated (GB)": 94.39,
      "memory_allocated (GB)": 51.7,
      "rewards/accuracies": 0.69140625,
      "rewards/chosen": -0.150390625,
      "rewards/margins": 0.8046875,
      "rewards/rejected": -0.95703125,
      "step": 20,
      "total_memory_available (GB)": 94.62
    },
    {
      "epoch": 0.15,
      "grad_norm": 1.9487425088882446,
      "learning_rate": 0.00015,
      "logits/chosen": -2.859375,
      "logits/rejected": -2.84375,
      "logps/chosen": -294.0,
      "logps/rejected": -394.0,
      "loss": 0.2394,
      "max_memory_allocated (GB)": 94.39,
      "memory_allocated (GB)": 51.7,
      "rewards/accuracies": 0.90625,
      "rewards/chosen": -1.984375,
      "rewards/margins": 4.8125,
      "rewards/rejected": -6.8125,
      "step": 30,
      "total_memory_available (GB)": 94.62
    },
    {
      "epoch": 0.21,
      "grad_norm": 1.0650818347930908,
      "learning_rate": 0.0002,
      "logits/chosen": -2.875,
      "logits/rejected": -2.828125,
      "logps/chosen": -284.0,
      "logps/rejected": -388.0,
      "loss": 0.1181,
      "max_memory_allocated (GB)": 94.39,
      "memory_allocated (GB)": 51.7,
      "rewards/accuracies": 0.95703125,
      "rewards/chosen": -0.62109375,
      "rewards/margins": 5.0625,
      "rewards/rejected": -5.6875,
      "step": 40,
      "total_memory_available (GB)": 94.62
    },
    {
      "epoch": 0.26,
      "grad_norm": 0.5032941699028015,
      "learning_rate": 0.00025,
      "logits/chosen": -2.890625,
      "logits/rejected": -2.875,
      "logps/chosen": -278.0,
      "logps/rejected": -412.0,
      "loss": 0.0583,
      "max_memory_allocated (GB)": 94.39,
      "memory_allocated (GB)": 51.7,
      "rewards/accuracies": 0.984375,
      "rewards/chosen": -0.8828125,
      "rewards/margins": 6.96875,
      "rewards/rejected": -7.84375,
      "step": 50,
      "total_memory_available (GB)": 94.62
    }
  ],
  "logging_steps": 10,
  "max_steps": 1000,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 6,
  "save_steps": 10,
  "total_flos": 4.19960425807872e+17,
  "train_batch_size": 8,
  "trial_name": null,
  "trial_params": null
}