vllm-ascend

vllm-ascend is a community-maintained add-on that connects vLLM to Huawei Ascend NPU hardware [GitHub description:1; README.md:53]. ^[1a][1b] It provides a hardware-pluggable interface that separates Ascend NPU integration from vLLM [README.md:55]. ^[2] Through it, supported open-source models, including Transformer-like, Mixture-of-Experts, embedding, and multi-modal LLMs, can run on Ascend NPU [README.md:57]. ^[3]