ontocord/RedPajama3b_v1-autoredteam_helpfulness-train

{
    "method": "orpo",
    "dataset": "autoredteam",
    "model": "togethercomputer/RedPajama-INCITE-Base-3B-v1",
    "tokenizer": "togethercomputer/RedPajama-INCITE-Instruct-3B-v1",
    "train_data_path": "/leonardo_work/EUHPC_E03_068/datasets/working/autoredteam_helpfulness_v1-train.jsonl",
    "test_data_path": "/leonardo_work/EUHPC_E03_068/datasets/working/autoredteam_helpfulness_v1-test.jsonl",
    "lr": 0.01,
    "train_batch_size": 600,
    "eval_batch_size": 600,
    "num_epochs": 1,
    "seed": 42,
    "eval_only": False,
    "evaluation_size": None,
    "gradient_accumulation_steps": 5,
    "checkpoint_path": None,
    "experiment_name": "RedPajama3b_v1-autoredteam_helpfulness-train",
    "experiment_group": "results",
    "reference_model": "togethercomputer/RedPajama-INCITE-Instruct-3B-v1",
    "context_length": 1024,
    "train_summarization": "",
    "dpo_beta": 0.1,
    "orpo_beta": 0.1,
    "kl_coef": 0.0,
    "reward_model": "",
    "bestofn_size": 4,
    "train_reward_model": "",
}