unslothai/unsloth · error · HTTPException

unable to resolve seed dataset path

Error message

unable to resolve seed dataset path

What it means

HTTP 422 raised in POST /seed/inspect when data_files were supplied but _resolve_seed_hf_path(dataset_name, data_files, split) could not map them to a concrete path pattern. The resolved path (e.g. datasets/{name}/**/*.parquet or a specific file glob) is what downstream recipe execution reads, so an unresolvable mapping aborts the inspect.

Source

Thrown at studio/backend/routes/data_recipe/seed.py:388

            raise log_and_http_error(
                exc,
                422,
                "seed inspect failed",
                event = "data_recipe.seed.hf_preview_failed",
                log = logger,
            ) from exc

    if not preview_rows:
        raise HTTPException(status_code = 422, detail = "dataset appears empty or unreadable")
    preview_rows = _serialize_preview_rows(preview_rows)
    columns = _extract_columns(preview_rows)

    if not data_files:
        resolved_path = f"datasets/{dataset_name}/**/*.parquet"
    else:
        resolved_path = _resolve_seed_hf_path(dataset_name, data_files, split)
        if not resolved_path:
            raise HTTPException(status_code = 422, detail = "unable to resolve seed dataset path")

    return SeedInspectResponse(
        dataset_name = dataset_name,
        resolved_path = resolved_path,
        columns = columns,
        preview_rows = preview_rows,
        split = split,
        subset = subset,
    )


def _extract_text_from_file(file_path: Path, ext: str) -> str:
    """Extract text from an uploaded file by extension, to markdown where possible."""
    if ext in {".txt", ".md"}:
        raw = file_path.read_text(encoding = "utf-8", errors = "ignore")
    elif ext == ".pdf":
        import pymupdf4llm
        raw = pymupdf4llm.to_markdown(

View on GitHub (pinned to 203007d190)

Solutions

1. List the repo files on huggingface.co (Files tab) and pass exact, current data_files paths.
2. Drop data_files and let the default 'datasets/{name}/**/*.parquet' resolution apply.
3. Verify the split argument matches the split directories embedded in the file paths.

Example fix

# before
inspect(dataset_name='org/repo', split='train', data_files=['data/train-0001-of-0002.parquet'])

# after
inspect(dataset_name='org/repo', split='train', data_files=['data/train-00000-of-00002.parquet'])  # exact current filename

// Verify data_files exist in the repo before inspect
const tree = await fetch(`https://huggingface.co/api/datasets/${encodeURIComponent(name)}/tree/main`).then(r => r.json());
const paths = new Set(tree.map(f => f.path));
const missing = dataFiles.filter(f => !paths.has(f));
if (missing.length) throw new Error(`not in repo: ${missing.join(', ')}`);

On 422 'unable to resolve seed dataset path', fall back to calling inspect without data_files; if that succeeds, the file list was wrong — refresh it from the repo.

When it happens

Trigger: POST /seed/inspect with data_files entries whose names do not exist in the repo, or whose layout does not let the resolver correlate data_files + split to a repo path (wrong split segment, renamed files, case mismatch).

Common situations: User copies data_files names from an older dataset revision; repo restructured (files moved/m renamed); split name embedded in the file path differs from the split argument.

Related errors

• dataset appears empty or unreadable
• hf_dataset is too long (max 256 chars)
• hf_dataset contains invalid characters or path segments
• subset is too long (max {MAX_HF_DATASET_OPTION_LENGTH} chars
• dataset_name must be a Hugging Face repo id like org/repo

AI-assisted analysis of unslothai/unsloth@203007d190 (2026-08-15). Data as JSON: /api/errors/3b0c4b7935a8b948. Report an issue: GitHub.