PyPI page
Home page
Author:
Rajveer Rathod
Summary:
Run longer contexts in less memory on Apple Silicon: KV-cache compression for MLX with Metal kernels (43 methods, from TurboQuant and KIVI to A2ATS)
Latest version:
0.94.9
Required dependencies:
mlx
|
mlx-lm
|
numpy
|
scipy
Optional dependencies:
build
|
furo
|
gitpython
|
matplotlib
|
mlx-vlm
|
mypy
|
myst-parser
|
pre-commit
|
psutil
|
pytest
|
pytest-cov
|
pytest-xdist
|
python-semantic-release
|
ruff
|
sphinx
|
sphinx-autodoc-typehints
|
twine
|
veloxquant-mlx
Downloads last day:
305
Downloads last week:
1,512
Downloads last month:
14,138
7-day smoothing