mixle.represent.segment module¶
Segmenters – cut a raw object of any modality into units, WITHOUT committing to a vocabulary.
A Segmenter turns one raw object (a string, an image, a waveform, a set of node features) into an array of
units: (n_units,) integer ids for a discrete alphabet, or (n_units, feat...) float features for a
continuous modality. That is the whole of “the tokenizer” that is not objective-dependent – it is a
decomposition, not a vocabulary. Discreteness (mapping units to a codebook of ids) is a separate, optional,
learned step (mixle.represent.quantize), so a segmenter never has to guess the right tokens.
Fixed segmenters (bytes, characters, patches, windows, whole-object, element-set) commit to nothing beyond where to cut and keep all information; the model learns the rest. Learned segmenters (a segmental HMM over boundaries) are the objective-coupled upgrade and plug into the same contract.
- class Segmenter[source]
Bases:
objectBase:
segment(raw) -> np.ndarray.discretesays whether units are ids (vs. float features).
- class ByteSegmenter[source]
Bases:
SegmenterA string/bytes ->
(n,)byte ids in[0, 256). The vocabulary-free text decomposition.
- class ElementSegmenter(alphabet)[source]
Bases:
SegmenterA sequence of hashable symbols (chars, amino acids, k-mers, categories) ->
(n,)ids via a fixed alphabet.Given
alphabet(the ordered symbols), each element maps to its index. Out-of-vocabulary symbols map tounknown_id, a state reserved past the declared alphabet (len(alphabet)), sonum_categoriesislen(alphabet) + 1. Unknown symbols used to collapse onto id0, which is already the first declared real category – an unseen residue/character/class then became positive evidence for a genuinely observed symbol. The alphabet must also be unique: a repeated entry used to overwrite its own index while still inflatingnum_categories, leaving an id no symbol could ever produce.The natural decomposition for proteins/genomes/any categorical sequence, and for characters (
alphabet=list(...)).- Parameters:
alphabet (list[Any])
- class PatchSegmenter(patch=8)[source]
Bases:
SegmenterAn image
(H, W)or(C, H, W)->(n_patches, patch_features)float units (ViT-style, no vocab).- Parameters:
patch (int)
- segment(raw)[source]
Split an image tensor into flattened non-overlapping patches.
- class WindowSegmenter(window=64, hop=None)[source]
Bases:
SegmenterA 1-D signal
(T,)->(n_frames, window)float units by a sliding window (seismic/audio/time-series).- segment(raw)[source]
Split a one-dimensional signal into fixed windows.
A signal shorter than
windowstill yields exactly one window (WindowSegmenter’s contract is to always return at least one unit, unlike PatchSegmenter which has no such contract and rejects a too-small image outright): the real samples fill[:len(x)]and only the remaining slots are zero-padded, so the padding never overwrites real data.
- class WholeSegmenter[source]
Bases:
SegmenterA single feature vector ->
(1, feat): the object is one unit (a pooled structure descriptor, a record).
- class SetSegmenter[source]
Bases:
SegmenterA set/list of feature vectors ->
(n_elements, feat): nodes of a graph, atoms of a molecule, taxa of a section.The general structured-object decomposition – a scientific structure becomes its set of element features (which a downstream model can further couple with a structure/message-passing embedding).