Factual package intelligence from PyPI

embedding-condensation 2.0 package intelligence

Measure layer-wise token embedding cosine similarity, assessing the severity of embedding condensation. Concept from [ICML 2026] Dispersion loss counteracts embedding condensation and improves generalization in small language models.

pip install embedding-condensation

Current version
2.0
Python requirement
>=3.9
License
Not declared
Distribution type
Pure Python wheel
Release files
2 (1 wheels, 1 source)
Download size
14.0 KiB
Release history
4 releases with files
Median release cadence
0 days

embedding-condensation dependencies

PyPI declares 8 unique dependency rules for this release. Environment markers are shown when supplied by the project.

  1. datasets >=2.14
  2. matplotlib >=3.7
  3. nltk >=3.8
  4. numpy >=1.24
  5. torch >=2.0
  6. tqdm >=4.65
  7. transformers <4.48,>=4.40

Compatibility and release files

embedding-condensation publishes 1 wheel and 1 source archive for version 2.0. Wheel platform tags: any.

No version-specific Python classifiers are declared.

Metadata completeness

PyPI does not currently declare: license. PyDeps marks these fields as unknown instead of guessing values.

Release activity and sources

PyPI lists 4 releases with files. The first dated release is ; 4 releases fall within the 365 days preceding the latest dated release. The current release files were uploaded on . The preceding dated release was .