{
    "model_name_or_path": "PKU-Alignment/alpaca-8b-reproduced-llama-3",
    "max_length": 512,
    "trust_remote_code": true,
    "train_datasets": [
        [
            "PKU-SafeRLHF/train",
            {
                "proportion": 1.0
            }
        ]
    ],
    "eval_datasets": [
        [
            "PKU-SafeRLHF/test",
            {
                "proportion": 1.0
            }
        ]
    ],
    "loss_type": "sequence-wise",
    "epochs": 2,
    "per_device_train_batch_size": 16,
    "per_device_eval_batch_size": 16,
    "gradient_accumulation_steps": 1,
    "gradient_checkpointing": true,
    "regularization": 0.001,
    "normalize_score_during_training": false,
    "normalizer_type": "ExponentialMovingAverage",
    "normalizer_momentum": 0.9,
    "lr": 2e-05,
    "lr_scheduler_type": "cosine",
    "lr_warmup_ratio": 0.03,
    "weight_decay": 0.1,
    "seed": 42,
    "fp16": true,
    "bf16": false,
    "tf32": null,
    "eval_strategy": "epoch",
    "eval_interval": 1000000,
    "need_eval": true,
    "eval_split_ratio": null,
    "output_dir": "/home/rwliang/Reinforcement-Learning/FINAL_PROJECT/rl-final/safe-rlhf/output/rm_llama3_8b",
    "log_type": "wandb",
    "log_dir": "/home/rwliang/Reinforcement-Learning/FINAL_PROJECT/rl-final/safe-rlhf/output/rm_llama3_8b",
    "log_project": "Safe-RLHF-RM",
    "log_run_name": "reward-2025-03-13-15-06-24",
    "save_16bit": true,
    "save_interval": 1000000,
    "local_rank": 0,
    "zero_stage": 3,
    "offload": "none",
    "deepspeed": false,
    "deepspeed_config": null,
    "deepscale": false,
    "deepscale_config": null,
    "global_rank": 0,
    "device": {
        "type": "torch.device",
        "repr": "device(type='cuda', index=0)"
    },
    "num_update_steps_per_epoch": 660,
    "total_training_steps": 1320
}