FAQs — vllm-ascend Currently, w8a8 quantization is already supported by vllm-ascend originally on v0.8.4rc2 or heigher, If you’re using vllm 0.7.3 version, w8a8 quantization is supporeted with the integration of vllm-ascend and mindie-turbo, please use pip install vllm-ascend[mindie-turbo] .