vllm-metax:0.19.0-maca.ai3.5.3.502-torch2.8-py310-ubuntu22.04镜像部署qwen3.5-35b-a3b-w8a8报错

在沐曦C500单卡使用vllm-metax:0.19.0-maca.ai3.5.3.502-torch2.8-py310-ubuntu22.04-amd64镜像部署qwen3.5-35b-a3b-w8a8报错，显示错误信息为ValueError: Tokenizer class TokenizersBackend does not exist or is not currently imported.在modelscope上查看到阿里发布的qwen3.5-35b-a3b的"tokenizer_class": "Qwen2Tokenizer",沐曦量化过后的显示为 "tokenizer_class": "TokenizersBackend",应该选择哪个版本的镜像进行部署？

尊敬的开发者您好，请参考developer.metax-tech.com/forum/t/fa-tie-qian-bi-kan-jing-xiang-shi-yong-wen-ti-ti-wen-mo-ban/267/ 详细描述