args.json

{
    "datasets": [
        "aditijb/collabllm-20q"
    ],
    "probs": [
        1
    ],
    "assistant_model_name": "meta-llama/Llama-3.2-1B-Instruct",
    "n_eval_per_dataset": 30,
    "max_prompt_length": 2048,
    "max_new_tokens": 256,
    "num_train_epochs": 1,
    "learning_rate": 5e-06,
    "save_total_limit": 10,
    "minimum_gap": 0.1,
    "per_device_train_batch_size": 4,
    "gradient_accumulation_steps": 8,
    "eval_steps": 1,
    "run_name": null,
    "resume_ckpt_dir": null,
    "output_dir": "./train/20q/dpo_train_offline",
    "seed": 42,
    "push_to_hub": true,
    "push_to_blob": false
}