microsoft/qlib · error · NotImplementedError

This type of input is not supported

Error message

This type of input is not supported

What it means

_prepare_seg in qlib/contrib/data/dataset.py converts a segment specifier into (start_date, end_date). It only understands a python slice, a 2-element list, or a 2-element tuple. Passing an int index, a numpy integer, a string segment name like 'train' (which top-level DatasetHK-style APIs use), or None raises NotImplementedError.

Source

Thrown at qlib/contrib/data/dataset.py:218

        # add memory (sample wise and daily)
        if self.memory_mode == "sample":
            self._memory = np.zeros((len(self._data), self.num_states), dtype=np.float32)
        elif self.memory_mode == "daily":
            self._memory = np.zeros((len(self._daily_index), self.num_states), dtype=np.float32)
        else:
            raise ValueError(f"invalid memory_mode `{self.memory_mode}`")

        # padding tensor
        self._zeros = np.zeros((self.seq_len, max(self.num_states, self._data.shape[1])), dtype=np.float32)

    def _prepare_seg(self, slc, **kwargs):
        fn = _get_date_parse_fn(self._index[0][1])
        if isinstance(slc, slice):
            start, stop = slc.start, slc.stop
        elif isinstance(slc, (list, tuple)):
            start, stop = slc
        else:
            raise NotImplementedError(f"This type of input is not supported")
        start_date = pd.Timestamp(fn(start))
        end_date = pd.Timestamp(fn(stop))
        obj = copy.copy(self)  # shallow copy
        # NOTE: Seriable will disable copy `self._data` so we manually assign them here
        obj._data = self._data  # reference (no copy)
        obj._label = self._label
        obj._index = self._index
        obj._memory = self._memory
        obj._zeros = self._zeros
        # update index for this batch
        date_index = self._index.get_level_values(1)
        obj._batch_slices = self._batch_slices[(date_index >= start_date) & (date_index <= end_date)]
        mask = (self._daily_index.values >= start_date) & (self._daily_index.values <= end_date)
        obj._daily_slices = self._daily_slices[mask]
        obj._daily_index = self._daily_index[mask]
        return obj

    def restore_index(self, index):

View on GitHub (pinned to 79633dd950)

Solutions

1. Pass an explicit range: dataset.prepare(slice(start, stop)) or dataset.prepare((start, stop)) with values convertible by _get_date_parse_fn.
2. Convert numpy scalars to python ints/strings before passing.
3. Do not pass segment-name strings; resolve them to concrete timestamps first if you reuse generic qlib workflow code.

Example fix

# before
seg = ds._prepare_seg(np.int64(100))  # -> NotImplementedError
# after
start, stop = ds._index[100][1], ds._index[-1][1]
seg = ds._prepare_seg((start, stop))  # 2-element tuple of timestamps

assert isinstance(seg, (slice, list, tuple)), f'segment must be slice or 2-element list/tuple, got {type(seg).__name__}'
if isinstance(seg, (list, tuple)):
    assert len(seg) == 2, 'segment container must have exactly two elements'

def is_valid_segment(s) -> bool:
    return isinstance(s, slice) or (isinstance(s, (list, tuple)) and len(s) == 2)

When it happens

Trigger: Calling dataset.prepare(segment) / _prepare_seg with a bare int or np.int64 (e.g. an index computed from enumerate), a single timestamp, or a named segment string that this RL dataset does not support.

Common situations: Mixing qlib's Dataset/DatasetH API conventions ('train'/'valid' segment names) with this RL dataset that expects explicit ranges; slicing with numpy scalar types produced by argmax/searchsorted.

Related errors

• invalid memory_mode `{self.memory_mode}`
• cannot assign data as `num_states==0`
• cannot clear memory as `num_states==0`
• This type of signal is not supported
• Must specify the path to save the dataset.

AI-assisted analysis of microsoft/qlib@79633dd950 (2026-08-15). Data as JSON: /api/errors/1bb08c066319ac19. Report an issue: GitHub.