vllm - 💡(How to fix) Fix [Doc]: VLLM_USE_FASTOKENS=1

📚 The doc issue

https://docs.vllm.ai/en/latest/configuration/optimization/#batch-level-dp-for-multi-modal-encoders states: Input Processing¶ fastokens Backend¶ By default vLLM uses the standard Hugging Face tokenizers library to power the fast tokenizer. For BPE tokenizers (Qwen, Llama, DeepSeek, GPT-OSS, etc.) you can switch to the fastokens Rust backend, a drop-in replacement that's substantially faster on encode/decode and on streaming detokenization. Enable it by setting VLLM_USE_FASTOKENS=1:

But there is no VLLM_USE_FASTOKENS in recent versions of vllm

Suggest a potential alternative/fix

No response

Before submitting a new issue...

• Make sure you already searched for relevant issues, and asked the chatbot living at the bottom right corner of the documentation page, which can answer lots of frequently asked questions.