Skip to content

knowledge(vllm-omni): 1 change(s) from intake #418

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
tzhouam merged 1 commit into from
Oct 25, 2026
Merged

knowledge(vllm-omni): 1 change(s) from intake #418

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
1 change: 1 addition & 0 deletions knowledge/repos/vllm-omni/components/serving/_index.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -64,3 +64,4 @@ sources: []
- [vLLM-Omni Serving：Seed-TTS 基准与准确率规则](rules-seed-tts.md)
- [vLLM-Omni Stage-CLI 测试规则](rules-stage-cli-tests.md)
- [vLLM-Omni Serving 规则 — MiniCPM-o 注意力后端对比](rules-minicpmo-attn-compare.md)
- [Diffusion speech 采样参数校验规则](rules-diffusion-speech-params.md)
19 changes: 19 additions & 0 deletions knowledge/repos/vllm-omni/components/serving/rules-diffusion-speech-params.md
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,19 @@
---
title: "Diffusion speech 采样参数校验规则"
created: 2026-10-25
updated: 2026-10-25
type: rule
tags: [vllm-omni, components, serving]
sources: ["PR #8685"]
---

# Diffusion speech 采样参数校验规则

## VLLM-OMNI-PR8685-SAMPLING-VALIDATION — diffusion speech 采样覆盖必须在请求边界做整值与有限性校验

- 触发：修改 `vllm_omni/entrypoints/openai/serving_speech.py` 中 `_create_diffusion_speech` 对 `extra_params`（`num_inference_steps`、`guidance_scale`）的解析，或 pure-diffusion `/v1/audio/speech` 的采样参数覆盖与错误映射。
- 强制：在调用 `generate()` 前完成全部校验，并经既有 ValueError→HTTP 400 通道返回结构化错误：`num_inference_steps` 必须是正整数，接受整数字符串与整值浮点；`guidance_scale` 必须是有限数，有限数值字符串、0 和负值合法；两参数都拒绝布尔值。类型转换的 `TypeError`、`ValueError`、`OverflowError` 均映射为对应 400 消息，不得外溢为 500。覆盖只写入 per-request sampling 对象，engine 默认参数保持不变。
- 禁止：静默截断非整值步数（12.5→12）；把布尔值当数字接受；接受 "NaN"/"Infinity" 字符串或 nan/±inf 浮点；让 `OverflowError`（如 10**400）穿过校验变成 500；在校验路径修改 `default_sampling_params_list` 中的默认对象。
- 验收：`tests/entrypoints/openai_api/test_serving_speech.py` 参数化覆盖非法输入（0、-1、0.5、12.5、True/False、nan/±inf、"NaN"/"Infinity"、10**400），断言对应 400 消息且 mock engine 的 `generate` 未被调用；合法用例（空参数取默认、"12"/12.0 整值、guidance "7.0"/0.0/-1.0）断言 200、`sampling_params_list` 中的值精确匹配且默认对象未被改动。^[PR #8685]

<!-- kb:rule status=active since=v0.30.0 -->
Loading