hiyouga/LlamaFactory · error · ValueError

File types should be identical.

Error message

File types should be identical.

What it means

Raised when a dataset loaded from a directory contains files with differing extensions: the type inferred from the first file is compared against every other file, and any mismatch aborts loading. LlamaFactory can only merge local files of a single format per dataset entry.

Source

Thrown at src/llamafactory/data/loader.py:89

        data_path = dataset_attr.dataset_name

    elif dataset_attr.load_from == "file":
        data_files = []
        local_path = os.path.join(data_args.dataset_dir, dataset_attr.dataset_name)
        if os.path.isdir(local_path):  # is directory
            for file_name in os.listdir(local_path):
                data_files.append(os.path.join(local_path, file_name))
        elif os.path.isfile(local_path):  # is file
            data_files.append(local_path)
        else:
            raise ValueError(f"File {local_path} not found.")

        data_path = FILEEXT2TYPE.get(os.path.splitext(data_files[0])[-1][1:], None)
        if data_path is None:
            raise ValueError("Allowed file types: {}.".format(",".join(FILEEXT2TYPE.keys())))

        if any(data_path != FILEEXT2TYPE.get(os.path.splitext(data_file)[-1][1:], None) for data_file in data_files):
            raise ValueError("File types should be identical.")
    else:
        raise NotImplementedError(f"Unknown load type: {dataset_attr.load_from}.")

    if dataset_attr.load_from == "ms_hub":
        check_version("modelscope>=1.14.0", mandatory=True)
        from modelscope import MsDataset  # type: ignore
        from modelscope.utils.config_ds import MS_DATASETS_CACHE  # type: ignore

        cache_dir = model_args.cache_dir or MS_DATASETS_CACHE
        dataset = MsDataset.load(
            dataset_name=data_path,
            subset_name=data_name,
            data_dir=data_dir,
            data_files=data_files,
            split=dataset_attr.split,
            cache_dir=cache_dir,
            token=model_args.ms_hub_token,
            use_streaming=data_args.streaming,

View on GitHub (pinned to f28afaf635)

Solutions

1. Make every file in the directory the same format/extension; convert the odd ones out (e.g. rewrite .json shards as .jsonl).
2. Remove non-data files from the directory (.DS_Store, README, backups).
3. Alternatively split the formats into separate directories and register each as its own dataset entry, then list both in the dataset field.
4. Point dataset_name at a single file instead of the directory if only one file is needed.

Example fix

# before: data/mydir/ contains part1.json, part2.jsonl

# after: data/mydir/ contains part1.jsonl, part2.jsonl
# (convert part1.json: one JSON object per line, then rename)

import os

def dir_uniform(d: str) -> bool:
    exts = {os.path.splitext(f)[-1] for f in os.listdir(d)}
    return len(exts) == 1

When it happens

Trigger: Pointing dataset_name at a directory that mixes .json and .jsonl, or .csv and .parquet, or contains stray files (e.g. .DS_Store, README.md, a lock file, or a backup copy with a different suffix).

Common situations: Incrementally exported data where a later export used a different format; macOS .DS_Store or editor backup files inside the data folder; combining sharded datasets saved with different serializers.

Related errors

• Allowed file types: {}.
• Unsupported model type: {getattr(config, 'model_type')}.
• Dataset converter {name} not found.
• Unknown mixing strategy: {data_args.mix_strategy}.
• Cannot specify `val_size` if `eval_dataset` is not None.

AI-assisted analysis of hiyouga/LlamaFactory@f28afaf635 (2026-08-14). Data as JSON: /api/errors/77f757d5725668a0. Report an issue: GitHub.