scripts/run_dpo.py [110:127]:
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
    )

    ##########################
    # Decontaminate benchmarks
    ##########################
    num_raw_train_samples = len(raw_datasets["train"])
    raw_datasets = raw_datasets.filter(
        decontaminate_humaneval,
        fn_kwargs={"text_column": "text_chosen"},
        batched=True,
        batch_size=10_000,
        num_proc=1,
        desc="Decontaminating HumanEval samples",
    )
    num_filtered_train_samples = num_raw_train_samples - len(raw_datasets["train"])
    logger.info(
        f"Decontaminated {num_filtered_train_samples} ({num_filtered_train_samples/num_raw_train_samples * 100:.2f}%) samples from the training set."
    )
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -



scripts/run_orpo.py [166:183]:
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
            )

    ##########################
    # Decontaminate benchmarks
    ##########################
    num_raw_train_samples = len(raw_datasets["train"])
    raw_datasets = raw_datasets.filter(
        decontaminate_humaneval,
        fn_kwargs={"text_column": "text_chosen"},
        batched=True,
        batch_size=10_000,
        num_proc=1,
        desc="Decontaminating HumanEval samples",
    )
    num_filtered_train_samples = num_raw_train_samples - len(raw_datasets["train"])
    logger.info(
        f"Decontaminated {num_filtered_train_samples} ({num_filtered_train_samples/num_raw_train_samples * 100:.2f}%) samples from the training set."
    )
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -