brainmass.process_sequence

brainmass.process_sequence(data, mode='stack')

Stack a sequence of data items along a new dimension.

This is the inverse operation of slice_data - while slice_data reduces a dimension via aggregation, stack_data creates a new dimension by stacking multiple items together.

Returns:

• mode=’stack’: Array/dict/tuple/list with new dimension at dim

• mode=’last’/’first’: Single item (same as data[-1] or data[0])

• mode=’avg’/’mean’/’max’/’min’: Aggregated tensor/dict/tuple/list

• mode=callable: Result of applying callable to stacked data

For dicts/tuples/lists, aggregation is applied recursively to each element.

Return type:

Array | ndarray | bool | number | bool | int | float | complex | Quantity | Tuple | List | Dict | Sequence

Raises:

• ValueError – If data is empty (cannot infer structure/type) or if mode is an unknown string.

• TypeError – If sequence contains mixed or incompatible types, or if mode is not a string or callable.

Parameters:

• data (PyTree)

• mode (str | Callable)

Notes

• All elements in data must have the same type and structure

• NumPy arrays are automatically converted to float32 tensors

• Dictionary keys must match across all elements

• Tuple/list lengths must match across all elements

• Recursive: handles nested structures (e.g., dict of tuples)