huggingface/pytorch-image-models · error · ValueError

num_batches must be a positive integer when specified.

Error message

num_batches must be a positive integer when specified.

What it means

num_batches (optional cap/override on the number of batches per epoch) must, when provided, be a positive integer; None/non-positive/fractional values fail this check in the constructor.

Source

Thrown at timm/data/scheduled_sampler.py:69

            seed: int = 0,
            drop_last: bool = True,
            shuffle_schedule: bool = True,
            num_batches: Optional[int] = None,
            choice_schedule: str = 'constant',
            schedule_epochs: Optional[int] = None,
            schedule_spread: float = 0.65,
            schedule_random_mix: float = 0.1,
    ) -> None:
        if not hasattr(sampler, '__len__'):
            raise TypeError('ScheduledBatchSampler requires a sampler with a length.')
        if len(sampler) <= 0:
            raise ValueError('ScheduledBatchSampler requires a non-empty sampler.')
        if not batch_sizes:
            raise ValueError('batch_sizes must contain at least one value.')
        if any(int(batch_size) != batch_size or batch_size <= 0 for batch_size in batch_sizes):
            raise ValueError('All scheduled batch sizes must be positive integers.')
        if num_batches is not None and (int(num_batches) != num_batches or num_batches <= 0):
            raise ValueError('num_batches must be a positive integer when specified.')
        if choice_schedule not in ('constant', 'progressive'):
            raise ValueError("choice_schedule must be 'constant' or 'progressive'.")
        if choice_schedule == 'progressive':
            if len(batch_sizes) < 2:
                raise ValueError('A progressive schedule requires at least two choices.')
            if schedule_epochs is None or int(schedule_epochs) != schedule_epochs or schedule_epochs <= 0:
                raise ValueError('schedule_epochs must be a positive integer for a progressive schedule.')
            if schedule_spread < 0:
                raise ValueError('schedule_spread must be non-negative.')
            if not 0 <= schedule_random_mix <= 1:
                raise ValueError('schedule_random_mix must be between 0 and 1.')

        self.sampler = sampler
        self.batch_sizes = tuple(int(batch_size) for batch_size in batch_sizes)
        self.choice_weights = self._normalize_choice_weights(choice_weights)
        self._active_choices = tuple(
            choice_index
            for choice_index, choice_weight in enumerate(self.choice_weights)

View on GitHub (pinned to 9a5261e31b)

Solutions

  1. Omit num_batches (pass None) to let the sampler infer it from the dataset length.
  2. Compute it defensively: num_batches = max(1, round(steps)) or verify steps > 0 before passing.
  3. Fix the underlying steps-per-epoch calculation (check batch size vs dataset size).

Example fix

# before
steps = len(dataset) // batch_size  # 0 when dataset < batch_size
sched = ScheduledBatchSampler(s, batch_sizes=[b], num_batches=steps)

# after
steps = max(1, len(dataset) // batch_size)  # or omit num_batches entirely
sched = ScheduledBatchSampler(s, batch_sizes=[b], num_batches=steps)
Defensive patterns

Strategy: validation

Validate before calling

num_batches = None if steps is None else max(1, int(round(steps))) if steps and steps > 0 else None

Prevention

When it happens

Trigger: ScheduledBatchSampler(..., num_batches=0), num_batches=-5, num_batches=2.5, or num_batches computed as len(dataset)//0 (ZeroDivision caught earlier) / a float division result.

Common situations: Computing num_batches from steps-per-epoch math that yields 0 on a tiny dataset; config typos; passing math.floor results of negative numbers.

Related errors


AI-assisted analysis of huggingface/pytorch-image-models@9a5261e31b (2026-08-27). Data as JSON: /api/errors/fb65660962a96ee9. Report an issue: GitHub.