piwheels

turboquant-vllm

TurboQuant KV cache compression for vLLM — fused Triton kernels, 3.76x compression, 3.7x faster decode on RTX 4090

Installation

In a virtualenv (see these instructions if you need to create one):

pip3 install turboquant-vllm

Dependencies

• accelerate
• vllm
• molmo-utils
• scipy
• torch
• einops
• torchvision
• transformers

Releases

Version	Released	Bullseye Python 3.9	Bookworm Python 3.11	Trixie Python 3.13	Files
1.5.0	2026-04-08
turboquant_vllm-1.5.0-py3-none-any.whl (86 KB)
1.4.1	2026-04-04
turboquant_vllm-1.4.1-py3-none-any.whl (83 KB)
1.4.0	2026-04-01
turboquant_vllm-1.4.0-py3-none-any.whl (82 KB)
1.3.0	2026-03-31
turboquant_vllm-1.3.0-py3-none-any.whl (80 KB)
1.2.2	2026-03-30
turboquant_vllm-1.2.2-py3-none-any.whl (76 KB)
1.2.1	2026-03-30
turboquant_vllm-1.2.1-py3-none-any.whl (75 KB)
1.2.0	2026-03-29
turboquant_vllm-1.2.0-py3-none-any.whl (74 KB)
1.1.1	2026-03-28
turboquant_vllm-1.1.1-py3-none-any.whl (58 KB)
1.1.0	2026-03-27
turboquant_vllm-1.1.0-py3-none-any.whl (58 KB)
1.0.0	2026-03-27
turboquant_vllm-1.0.0-py3-none-any.whl (58 KB)
0.1.0	2026-03-27
turboquant_vllm-0.1.0-py3-none-any.whl (59 KB)

Issues with this package?

• Search issues for this package
• Package or version missing? Open a new issue
• Something else? Open a new issue