vllm-project/vllm-omni

[Feature]: Rename ar_stop_token_ids to a more descriptive name

Open

#4,007 opened on May 30, 2026

View on GitHub
 (2 comments) (0 reactions) (1 assignee)Python (1,067 forks)github user discovery
good first issuehelp wanted

Repository metrics

Stars
 (4,990 stars)
PR merge metrics
 (PR metrics pending)

Description

🚀 The feature, motivation and pitch

The variable ar_stop_token_ids in vllm_omni/entrypoints/openai/serving_chat.py:2296 has an awkward name. It was introduced in PR #3896 and noted during code review (https://github.com/vllm-project/vllm-omni/pull/3896#discussion_r3328486645).

The name ar_stop_token_ids should be renamed to something more descriptive and idiomatic to improve code readability.

Alternatives

  • ar_stop_token_idsautoregressive_stop_token_ids (more explicit)
  • ar_stop_token_idsstop_token_ids_for_ar_stage (more descriptive of the context)
  • ar_stop_token_idsar_stage_stop_token_ids (alternative ordering)

Additional context

This variable is used in _build_multistage_generation_inputs to store stop token IDs resolved for the AR (autoregressive/LLM) stage. It is later assigned to default_stage_params.stop_token_ids when stage_type == "llm". A clearer name would help future readers.

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.

Contributor guide