unslothai/unsloth · error · ValueError

subset is too long (max {MAX_HF_DATASET_OPTION_LENGTH} chars

Error message

subset is too long (max {MAX_HF_DATASET_OPTION_LENGTH} chars)

What it means

Raised by the _check_subset field validator on TrainingStartRequest: the optional dataset subset/config name is capped at MAX_HF_DATASET_OPTION_LENGTH characters after trimming (empty-after-trim normalizes to None). Subsets are short identifiers like 'default' or 'plain_text', so an over-long value signals a pasted blob or wrong field mapping. It fires before the character-validity check.

Source

Thrown at studio/backend/models/training.py:249

            return None
        if len(v) > 256:
            raise ValueError("hf_dataset is too long (max 256 chars)")
        if ".." in v:
            raise ValueError("hf_dataset must not contain '..'")
        if any(_HF_DATASET_ID_SEGMENT_RE.fullmatch(segment) is None for segment in v.split("/")):
            raise ValueError("hf_dataset contains invalid characters or path segments")
        return v

    @field_validator("subset")
    @classmethod
    def _check_subset(cls, v: Optional[str]) -> Optional[str]:
        if v is None:
            return v
        v = v.strip()
        if not v:
            return None
        if len(v) > MAX_HF_DATASET_OPTION_LENGTH:
            raise ValueError(f"subset is too long (max {MAX_HF_DATASET_OPTION_LENGTH} chars)")
        if not valid_hf_dataset_config_name(v):
            raise ValueError("subset contains invalid characters")
        return v

    @field_validator(
        "model_local_path",
        "dataset_local_path",
        "model_snapshot_path",
        "dataset_snapshot_path",
    )
    @classmethod
    def _check_cache_local_path(cls, v: Optional[str]) -> Optional[str]:
        if v is None:
            return v
        v = v.strip()
        if not v:
            return None
        if len(v) > 4096:

View on GitHub (pinned to 203007d190)

Solutions

1. Send the short subset/config name only, e.g. 'default'.
2. Check your request mapping: long config blobs belong elsewhere, not in subset.
3. Omit subset entirely (it normalizes to None) if the dataset has a single config.

Example fix

// before
{ hf_dataset: "user/data", subset: JSON.stringify(config) }
// after
{ hf_dataset: "user/data", subset: "default" }

def subset_short_enough(body: dict, max_len: int) -> bool:
    v = (body.get("subset") or "").strip()
    return len(v) <= max_len

function subsetOk(subset: string | undefined, maxLen: number): boolean {
  return (subset?.trim().length ?? 0) <= maxLen;
}

When it happens

Trigger: POST a training start request with a subset string longer than the configured max (MAX_HF_DATASET_OPTION_LENGTH), e.g. a full config JSON accidentally placed in subset.

Common situations: Mapping the wrong config field (e.g. training args JSON) into subset; pasting a dataset README; a UI text-area instead of a short input bound to subset.

Related errors

• hf_dataset is too long (max 256 chars)
• hf_dataset contains invalid characters or path segments
• subset contains invalid characters
• dataset_slice_end must be greater than or equal to dataset_s
• split name is too long (max {MAX_HF_DATASET_OPTION_LENGTH} c

AI-assisted analysis of unslothai/unsloth@203007d190 (2026-08-15). Data as JSON: /api/errors/95a054e3112fc2f1. Report an issue: GitHub.