inference/research
verifiedhigh · 0.90updated 2026-10-03

vLLM Ascend plugin GitHub repository

repo

Fields

id
sources/niche-vllm-ascend-repo
Publisher
vLLM project (community)
Kind
repo
Accessed
2026-10-03
notes
vllm-ascend README: community-maintained hardware plugin for running vLLM on Huawei Ascend NPU, created by the vLLM community in 2025-02 following vLLM's [RFC] Hardware pluggable. Pinned stack, which is the whole story for this record: Python >= 3.10 and < 3.13, CANN == 9.1.0, PyTorch == 2.10.0, TorchNPU == 2.10.0.post4, vLLM at the SAME version as vllm-ascend, Linux only. Hardware listed: Atlas 800I A2/A3 Inference, Atlas A2/A3 Training, Atlas 300I Duo (Experimental). API metadata 2026-10-03: last push 2026-10-02, v0.23.0.post1 (2026-09-22), 4415 open issues, Apache-2.0.

Notes

vllm-ascend README: community-maintained hardware plugin for running vLLM on Huawei Ascend NPU, created by the vLLM community in 2025-02 following vLLM's [RFC] Hardware pluggable. Pinned stack, which is the whole story for this record: Python >= 3.10 and < 3.13, CANN == 9.1.0, PyTorch == 2.10.0, TorchNPU == 2.10.0.post4, vLLM at the SAME version as vllm-ascend, Linux only. Hardware listed: Atlas 800I A2/A3 Inference, Atlas A2/A3 Training, Atlas 300I Duo (Experimental). API metadata 2026-10-03: last push 2026-10-02, v0.23.0.post1 (2026-09-22), 4415 open issues, Apache-2.0.

Cross-references

This record references no other records.