PyPI page
Home page
Author:
fused-turboquant Contributors
Summary:
Fused Triton encode/decode kernels for TurboQuant KV cache compression, powered by Randomized Hadamard Transform.
Latest version:
0.1.0
Required dependencies:
numpy
|
scipy
|
torch
Optional dependencies:
accelerate
|
datasets
|
pytest
|
pytest-benchmark
|
ruff
|
transformers
|
triton
|
triton-windows
|
vllm
Downloads last day:
1
Downloads last week:
9
Downloads last month:
10
7-day smoothing