目录

vLLM Ascend Parameter Adaptation

This submission adapts --attention-config and --ray-workers-use-nsight for Ascend NPU.

Source-level delivery is included in:

attention-config-ray-workers-use-nsight/

That directory contains changed vllm_ascend source files, focused unit tests, user documentation, and the parameter-specific SKILL.md.

The same implementation is also provided as a unified patch under patches/, with Ascend 910B2C validation records in verification/.

Validation summary:

attention_config or nsight focused tests: 5 passed
full tests/ut/test_platform.py: 57 passed, 1 skipped
关于

本仓库用于CCF开源创新大赛

55.3 MB
邀请码
    Gitlink(确实开源)
  • 加入我们
  • 官网邮箱:gitlink@ccf.org.cn
  • QQ群
  • QQ群
  • 公众号
  • 公众号

版权所有:中国计算机学会技术支持:开源发展技术委员会
京ICP备13000930号-9 京公网安备 11010802047560号