huggingface/open-r1 · error

Either `dataset_name` or `dataset_mixture` must be provided

Error message

Either `dataset_name` or `dataset_mixture` must be provided

What it means

The training script's dataset config (DatasetConfig.__post_init__) requires that you identify the training data either via dataset_name (a single HF Hub dataset) or dataset_mixture (a multi-dataset mixture dict). If both are None, the dataclass immediately raises this ValueError at config-construction time, because the trainer would otherwise have no data to load.

Source

Thrown at src/open_r1/configs.py:80

                      - col1
                      - col2
                    weight: 0.5
                seed: 42
                test_split_size: 0.1
    """

    # Override the dataset_name to make it optional
    dataset_name: Optional[str] = field(
        default=None, metadata={"help": "Dataset name. Can be omitted if using dataset_mixture."}
    )
    dataset_mixture: Optional[dict[str, Any]] = field(
        default=None,
        metadata={"help": "Configuration for creating dataset mixtures with advanced options like shuffling."},
    )

    def __post_init__(self):
        if self.dataset_name is None and self.dataset_mixture is None:
            raise ValueError("Either `dataset_name` or `dataset_mixture` must be provided")

        if self.dataset_mixture is not None:
            if not isinstance(self.dataset_mixture, dict) or "datasets" not in self.dataset_mixture:
                raise ValueError(
                    "dataset_mixture must be a dictionary with a 'datasets' key. "
                    "Expected format: {'datasets': [...], 'seed': int}"
                )

            datasets_list = []
            datasets_data = self.dataset_mixture.get("datasets", [])

            if isinstance(datasets_data, list):
                for dataset_config in datasets_data:
                    datasets_list.append(
                        DatasetConfig(
                            id=dataset_config.get("id"),
                            config=dataset_config.get("config"),
                            split=dataset_config.get("split", "train"),

View on GitHub (pinned to 1416fa0cf2)

Solutions

1. Pass a dataset name: dataset_name="HuggingFaceH4/ultrachat_200k" (or --dataset_name on the CLI).
2. Or provide a mixture dict: dataset_mixture={"datasets": [{"id": "ds1"}, {"id": "ds2"}], "seed": 42}.
3. Check your YAML/JSON/CLI arg spelling so the value actually populates dataset_name/dataset_mixture.

Example fix

// before
args = SFTConfig(output_dir="out")  # dataset_name=None, dataset_mixture=None -> ValueError
// after
args = SFTConfig(output_dir="out", dataset_name="HuggingFaceH4/ultrachat_200k")

cfg = DatasetConfig(dataset_name=args.get("dataset_name"), dataset_mixture=args.get("dataset_mixture"))
if cfg.dataset_name is None and cfg.dataset_mixture is None:
    raise SystemExit("Set --dataset_name or --dataset_mixture before launching training")

def has_dataset(cfg) -> bool:
    return getattr(cfg, "dataset_name", None) is not None or getattr(cfg, "dataset_mixture", None) is not None

try:
    cfg = DatasetConfig(**raw_args)
except ValueError as e:
    if "dataset_name" in str(e) or "dataset_mixture" in str(e):
        sys.exit("Config error: provide --dataset_name or --dataset_mixture")
    raise

When it happens

Trigger: Instantiating the SFT/GRPO/DPO config dataclass (e.g. via HfArgumentParser from CLI args or in Python) with neither --dataset_name nor --dataset_mixture set.

Common situations: New training configs copied from templates where the dataset fields were deleted; CLI runs that omitted the dataset flag; YAML/JSON configs where the key is misspelled (e.g. dataset_names) so it lands nowhere; script defaults where dataset_name=None was left in place.

Related errors

• dataset_mixture must be a dictionary with a 'datasets' key.
• 'datasets' must be a list of dataset configurations
• Column names must be consistent across all dataset configura
• CF_TESTS_FOLDER environment variable not set! Please downloa
• Dataset Question Field Error: {prompt_column} is not support

AI-assisted analysis of huggingface/open-r1@1416fa0cf2 (2026-08-30). Data as JSON: /api/errors/90f27a571826d0f6. Report an issue: GitHub.