huggingface/transformers · error · ValueError

Unable to convert output '{key}' (type: {type(value).__name_

Error message

Unable to convert output '{key}' (type: {type(value).__name__}) to tensor: {str(e)}
You can try:
  1. Use padding=True to ensure all outputs have the same shape
  2. Set return_tensors=None to return Python objects instead of tensors

What it means

The generic tensor-conversion failure in BatchFeature.convert_to_tensors: as_tensor(value) raised for a non-overflow key (ragged nested lists, mismatched lengths across the batch, unsupported element types). The error names the key, its python type, and the underlying message, and suggests padding or returning python objects.

Source

Thrown at src/transformers/feature_extraction_utils.py:206

            # Skip keys explicitly marked for no conversion
            if skip_tensor_conversion and key in skip_tensor_conversion:
                continue

            # Skip values that are not array-like
            if not _is_tensor_or_array_like(value):
                continue

            try:
                if not is_tensor(value):
                    tensor = as_tensor(value)
                    self[key] = tensor
            except Exception as e:
                if key == "overflowing_values":
                    raise ValueError(
                        f"Unable to create tensor for '{key}' with overflowing values of different lengths. "
                        f"Original error: {str(e)}"
                    ) from e
                raise ValueError(
                    f"Unable to convert output '{key}' (type: {type(value).__name__}) to tensor: {str(e)}\n"
                    f"You can try:\n"
                    f"  1. Use padding=True to ensure all outputs have the same shape\n"
                    f"  2. Set return_tensors=None to return Python objects instead of tensors"
                ) from e

        return self

    def to(self, *args, **kwargs) -> "BatchFeature":
        """
        Send all values to device by calling `v.to(*args, **kwargs)` (PyTorch only). This should support casting in
        different `dtypes` and sending the `BatchFeature` to a different `device`.

        Args:
            args (`Tuple`):
                Will be passed to the `to(...)` function of the tensors.
            kwargs (`Dict`, *optional*):
                Will be passed to the `to(...)` function of the tensors.

View on GitHub (pinned to a597f97485)

Solutions

  1. Add padding=True (plus max_length if a fixed size is needed) so all sequences align
  2. Use return_tensors=None when you need to keep ragged python/numpy structures
  3. Verify each key in the batch has a consistent shape before conversion

Example fix

# before
fe(list_of_variable_length_audio, return_tensors="pt")

# after
fe(list_of_variable_length_audio, return_tensors="pt", padding=True)
Defensive patterns

Strategy: fallback

Validate before calling

def lengths_uniform(values) -> bool:
    import numpy as np
    if isinstance(values, (list, tuple)) and values and isinstance(values[0], (list, tuple, np.ndarray)):
        return len({len(v) for v in values}) == 1
    return True

Try / catch

try:
    out = fe(batch, return_tensors="pt")
except ValueError as e:
    if "to tensor" in str(e):
        out = fe(batch, return_tensors="pt", padding=True)
    else:
        raise

Prevention

When it happens

Trigger: Calling a feature extractor with return_tensors='pt'/'np' on an unpadded ragged batch (sequences of different lengths), or a value that is a nested list of inconsistent shapes for a key other than overflowing_values.

Common situations: Batching variable-length audio without padding=True; mixing single example and batched arrays; passing strings or object arrays as feature values.

Related errors

AI-assisted analysis of huggingface/transformers@a597f97485 (2026-08-14). Data as JSON: /api/errors/c35fc9cb199568bf. Report an issue: GitHub.