Skip to main content
Pre-release

This release is a pre-release and may not be stable for production use.

xlite logo

Xlite轻量化推理运行时

介绍

xlite (GVirt前端):轻量级Transformer模型运行时,支持多样性算力协同,当前支持在昇腾硬件上高效运行。 xlite公开了Transformer运行所需的模型构图以及算子,所有算子基于昇腾AscendC/CCE开发。

背景与动机

在大模型推理场景中,传统的单流串行执行模式存在以下问题:

  • 核间负载不均:不同AICORE之间的任务分配不均衡,部分核心闲置
  • 资源浪费多:计算资源和传输资源利用率低,存在显著浪费
  • 执行时间长:Host CPU下发算子开销大,造成严重的host bond问题

xlite通过以下技术手段解决上述问题:

  • 多流并行:充分利用卡内资源,将单流串行改为多流并行执行
  • 核间负载均衡:核间负载均衡,提升资源利用率
  • CPU NPU协同:C++侧完全消除Python的GC、线程等干扰,简化Host tiling计算,去除小块内存申请释放及拷贝,消除Host bond

性能效果显著:

在GLM-4.7双机推理场景(40K输入、1K输出、prefix cache命中率约90%)下:

  • TPOT时延降低17%~30%
  • 吞吐提升13%~41%

详细性能数据参考 PR #7935。

软件架构

image

xlite已适配vllm_ascend,可通过xlite_graph_config配置快速使能xlite加速效果,使用方法参考官方指导文档。

xlite支持模型见模型列表。

快速开始

安装

# 安装vllm_ascend, 可参考https://github.com/vllm-project/vllm-ascend/blob/main/README.md
# 安装xlite
pip install xlite

离线推理示例

import os
from vllm import LLM

# xlite默认支持decode-only模式, 可通过设置 "full_mode": True 使能full模式
model = LLM(model="path/to/Qwen3-32B", tensor_parallel_size=8, additional_config={"xlite_graph_config": {"enabled": True, "full_mode": True}})
outputs = model.generate("Hello, how are you?")

在线服务示例

vllm serve path/to/Qwen3-32B --tensor-parallel-size 8 --additional-config='{"xlite_graph_config": {"enabled": true, "full_mode": true}}'

开发指南

编译构建、容器镜像、源码安装等开发相关内容,请参考 开发指南

环境变量

xlite使用的环境变量及其配置方法,请参考 环境变量

目录结构

.
├── csrc/    - 轻量化运行时的核心代码
├── xlite/   - python代码,包括tools等
├── doc/     - 相关文档介绍
├── docker/  - 容器镜像Dockerfile
└── tests/   - 测试用例

致谢

本项目的核心算子初始版本和思路来自于华为终端小艺AI Infra团队的贡献,相关优化实现可参考论文:

《XY-Serve: End-to-End Versatile Production Serving for Dynamic LLM Workloads》 [ASPLOS 2026]

Metadata

Release files for xlite 0.1.0rc9.dev210

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Built distributions (wheels)

Table of built distributions (wheels) for xlite 0.1.0rc9.dev210
File Actions
xlite-0.1.0rc9.dev210-cp312-cp312-manylinux2014_x86_64.whl CPython 3.12 CPython 3.12 Linux glibc 2.17+ x86-64 Details for xlite-0.1.0rc9.dev210-cp312-cp312-manylinux2014_x86_64.whl
xlite-0.1.0rc9.dev210-cp312-cp312-manylinux2014_aarch64.whl CPython 3.12 CPython 3.12 Linux glibc 2.17+ ARM64 Details for xlite-0.1.0rc9.dev210-cp312-cp312-manylinux2014_aarch64.whl
xlite-0.1.0rc9.dev210-cp311-cp311-manylinux2014_x86_64.whl CPython 3.11 CPython 3.11 Linux glibc 2.17+ x86-64 Details for xlite-0.1.0rc9.dev210-cp311-cp311-manylinux2014_x86_64.whl
xlite-0.1.0rc9.dev210-cp311-cp311-manylinux2014_aarch64.whl CPython 3.11 CPython 3.11 Linux glibc 2.17+ ARM64 Details for xlite-0.1.0rc9.dev210-cp311-cp311-manylinux2014_aarch64.whl

Total release size: 3.1 MB

Release files / xlite-0.1.0rc9.dev210-cp312-cp312-manylinux2014_x86_64.whl

Download URL xlite-0.1.0rc9.dev210-cp312-cp312-manylinux2014_x86_64.whl
Size 784.4 kB
Tags CPython 3.12 Linux glibc 2.17+ x86-64
SHA-256 checksum
How to use checksums
f77debc259fb706a9d1e253996e8ce2b5c0943b3c626cbe57638c3e2346e7986
BLAKE2b-256 checksum
How to use checksums
0eec6f099419dee996d2e2b00f007426c7ff8974539be32992f5afde53745c7c
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via twine/6.2.0 CPython/3.12.7

Release files / xlite-0.1.0rc9.dev210-cp312-cp312-manylinux2014_aarch64.whl

Download URL xlite-0.1.0rc9.dev210-cp312-cp312-manylinux2014_aarch64.whl
Size 744.2 kB
Tags CPython 3.12 Linux glibc 2.17+ ARM64
SHA-256 checksum
How to use checksums
26b15763232541018b042f9077e8d7ae0079de05695031b01895eb605dbc5f07
BLAKE2b-256 checksum
How to use checksums
0a3eaf4f44745d8a3c7657a1fb4a2ee463ce89557d1ea6e8fa9b85a4d91b1ad8
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via twine/6.2.0 CPython/3.12.7

Release files / xlite-0.1.0rc9.dev210-cp311-cp311-manylinux2014_x86_64.whl

Download URL xlite-0.1.0rc9.dev210-cp311-cp311-manylinux2014_x86_64.whl
Size 780.0 kB
Tags CPython 3.11 Linux glibc 2.17+ x86-64
SHA-256 checksum
How to use checksums
26caa721bc85f020987d929633f5ee58ee0043126d21c2b327acbbbf942d9222
BLAKE2b-256 checksum
How to use checksums
779f84b345b4fd21d4add49e3255308c7eb44bf0d8b462f92c8813027960d65a
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via twine/6.2.0 CPython/3.11.11

Release files / xlite-0.1.0rc9.dev210-cp311-cp311-manylinux2014_aarch64.whl

Download URL xlite-0.1.0rc9.dev210-cp311-cp311-manylinux2014_aarch64.whl
Size 743.6 kB
Tags CPython 3.11 Linux glibc 2.17+ ARM64
SHA-256 checksum
How to use checksums
b228633a76d411bbb7a3a0a00b91a0af5bd0ff29261aa4eb4a38c08093188f69
BLAKE2b-256 checksum
How to use checksums
87da8f4bd750bffeb5a650855a8bc567295fd3b55672820d092dc7175c2982bd
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via twine/6.2.0 CPython/3.11.11

Security for xlite 0.1.0rc9.dev210

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

No provenance found

None of the files in this release were uploaded with provenance metadata.

A note on package security

PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.

To learn more about protecting yourself from malicious actors, visit our package security documentation.

Release history Release notifications | RSS feed

This release
Anthropic, PBC Visionary sponsor Bloomberg Visionary sponsor Hudson River Trading Visionary sponsor Meta Visionary sponsor NVIDIA Visionary sponsor Microsoft Sustainability sponsor Depot Continuous Integration AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page