docs(vllm_guide): let vLLM pin torch/torchaudio in the installation steps

[qiulang]

对应 #3006 第 2 点。安装步骤去掉手动 pip install torch torchaudio,改为「按 nvidia-smi 的 CUDA 版本选 vLLM → 由 vLLM 钉定匹配的 torch/torchaudio/torchvision」,并说明三者 ABI 锁死、不应单独安装(附 vLLM requirements/cuda.txt 链接)。en / zh / zh_v2 三份同步修改。

多个 commit,合并时可 squash。

[gemini-code-assist]

Code Review

This pull request updates the installation guides (docs/vllm_guide.md, docs/vllm_guide_zh.md, and docs/vllm_guide_zh_v2.md) to clarify that vLLM should be installed first to automatically manage the matching, ABI-locked PyTorch, torchaudio, and torchvision dependencies, which helps prevent CUDA driver mismatch errors. The review feedback highlights two issues: the English guide accidentally omitted several required dependencies (such as safetensors and fastapi) that remain in the Chinese versions, and a markdown code block in the Chinese guide is missing the bash language identifier, which breaks syntax highlighting.

Important

The consumer version of Gemini Code Assist on GitHub is being sunset. Starting June 18, 2026, new organization installations will be blocked, and all code review activity will officially cease on July 17, 2026.
For more details on the timeline and next steps, please review the Help Documentation.

[gemini-code-assist]

在英文版指南中，移除了 safetensors tiktoken websockets regex fastapi uvicorn python-multipart 的安装步骤，但这与中文版指南（vllm_guide_zh.md 和 vllm_guide_zh_v2.md）不一致。这些依赖项对于后续章节中介绍的 vLLM 服务（如 serve_vllm.py 和 serve_realtime_ws.py）是必需的。建议在英文版中也保留这些依赖的安装。

Suggested change

	pip install funasr>=1.3.0
	pip install funasr>=1.3.0
	pip install safetensors tiktoken websockets regex fastapi uvicorn python-multipart

[gemini-code-assist]

这里代码块的开头缺少了 bash 语言标识符（变成了 ```），这会导致在渲染文档时无法正确进行语法高亮。建议修改为 ```bash，以保持与英文版及 zh_v2 版本的一致性。

[LauraGPT]

感谢 @qiulang!这版把「让 vLLM 钉定 torch/torchaudio/torchvision 三件套(ABI 锁死,不要手动单独装)」讲清楚了,正好对上 #3006 第 2 点反馈的驱动/CUDA 版本不匹配的坑。版本号也已核实(vllm==0.19.1 等真实存在,当前 latest 0.23.0),文中作为示例、可按 nvidia-smi 的驱动 CUDA 自行调整,表述合理。en/zh/zh_v2 三份同步,已 squash 合并。再次感谢贡献 🙏