[just test] tmp fix k3 in step0

[zoooo0820]

Motivation

💡 If this PR is a Cherry Pick, the PR title needs to follow the format by adding the [Cherry-Pick] label at the very beginning and appending the original PR ID at the end. For example, [Cherry-Pick][CI] Add check trigger and logic(#5191)

💡 如若此PR是Cherry Pick，PR标题需遵循格式，在最开始加上[Cherry-Pick]标签，以及最后面加上原PR ID，例如[Cherry-Pick][CI] Add check trigger and logic(#5191)

Modifications

Usage or Command

Accuracy Tests

Checklist

• Add at least a tag in the PR title.
  • Tag list: [[FDConfig],[APIServer],[Engine], [Scheduler], [PD Disaggregation], [Executor], [Graph Optimization], [Speculative Decoding], [RL], [Models], [Quantization], [Loader], [OP], [KVCache], [DataProcessor], [BugFix], [Docs], [CI], [Optimization], [Feature], [Benchmark], [Others], [XPU], [HPU], [GCU], [DCU], [Iluvatar], [Metax]]
  • You can add new tags based on the PR content, but the semantics must be clear.
• Format your code, run pre-commit before commit.
• Add unit tests. Please write the reason in this PR if no unit tests.
• Provide accuracy results.
• If the current PR is submitting to the release branch, make sure the PR has been submitted to the develop branch, then cherry-pick it to the release branch with the [Cherry-Pick] PR tag.

[paddle-bot]

Thanks for your contribution!

[PaddlePaddle-bot]

🤖 Paddle-CI-Agent | ci_status_monitor | 2026-05-22 22:29:21

CI报告基于以下代码生成（30分钟更新一次）:

• PR commit: 0159735
• Merge base: d7a9f9d (branch: develop)
• 查看完整 Diff
• CI 详情

---

1 任务总览

Required 任务已全部通过（10/10），当前 PR 的必选 CI 不阻塞合并，建议通过。另有 4 个 Optional 任务失败，仅供参考。

总执行（rerun次数）	总任务	✅ 通过	❌ 失败	⏳ 运行中	⏸️ 等待中	跳过
41(0)	41	37	4	0	0	0

2 任务状态汇总

2.1 Required任务 : 10/10 通过

必选任务阻塞合并，失败需优先处理。

状态	任务	耗时	根因	修复建议	日志	重跑
✅	其余 10 个必选任务通过	-	-	-	-	-

2.2 可选任务 — 27/31 通过

可选任务不阻塞合并，失败仅供参考。

状态	任务	耗时	日志	重跑
❌	Run iluvatar Tests / run_iluvatar_cases	1m25s	Job	-
❌	Check PR Template	9s	Job	-
❌	CI_HPU	1h4m	Job	-
❌	Trigger Jenkins for PR	7m40s	Job	-
✅	其余 27 个可选任务通过	-	-	-

---

3 失败详情（仅 required）

无 required 失败任务。

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.
⚠️ Please upload report for BASE (develop@d7a9f9d). Learn more about missing BASE report.

Additional details and impacted files

@@            Coverage Diff             @@
##             develop    #7897   +/-   ##
==========================================
  Coverage           ?   63.60%           
==========================================
  Files              ?      462           
  Lines              ?    64507           
  Branches           ?     9887           
==========================================
  Hits               ?    41030           
  Misses             ?    20701           
  Partials           ?     2776           

Flag	Coverage Δ
GPU	72.73% <100.00%> (?)
XPU	7.11% <0.00%> (?)

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[PaddlePaddle-bot]

🤖 Paddle-CI-Agent | pr_review | 2026-05-22 18:42:07

📋 Review 摘要

PR 概述：将 capture_model() 与 capture_block_wise_graphs() 移至 deterministic 模式 RNG/share_inputs 重置之后，修复 deterministic 模式下 CUDAGraph 捕获时使用 stale 状态导致的 k3 计算错误。
变更范围：fastdeploy/worker/gpu_worker.py（graph_optimize_and_warm_up_model）
影响面 Tag：[Graph Optimization]

问题

级别	文件	概述
🟡 建议	gpu_worker.py:255	其他硬件 Worker 是否需同步相同的 capture 顺序修复
📝 PR 规范	—	标题缺少官方 Tag；PR 描述各 Section 内容为空

📝 PR 规范检查

存在两处规范问题：

1. 标题 [just test] tmp fix k3 in step0 使用了非官方 Tag，且含临时性描述文字；
2. PR 描述中 Motivation / Modifications / Usage or Command / Accuracy Tests 均为空，Checklist 全未勾选。

标题建议（可直接复制）：

• [BugFix] Fix CUDAGraph capture order in deterministic mode to use clean RNG state

PR 描述建议（可直接复制）：

## Motivation
在 deterministic 模式下，`graph_optimize_and_warm_up_model` 先执行 `capture_model()` 再重置 RNG 和 `share_inputs`，导致 CUDAGraph 在 warmup 之后的 stale 状态（被消耗的 CUDA RNG、infer_seed/stop_flags/seq_lens 等旧数据）下完成捕获，进而引发 k3（step0）计算结果不确定。

## Modifications
- `fastdeploy/worker/gpu_worker.py`：将 `capture_model()` 和 `capture_block_wise_graphs()` 的调用位置从 deterministic 模式 RNG/share_inputs 重置块之前移至之后，确保 CUDAGraph 在干净初始状态下完成捕获。

## Usage or Command
N/A

## Accuracy Tests
N/A（此修复为执行顺序调整，不影响精度输出；deterministic 模式下输出一致性需在完整 CI 中验证）

## Checklist

- [x] Add at least a tag in the PR title.
  - Tag list: [`[FDConfig]`,`[APIServer]`,`[Engine]`, `[Scheduler]`, `[PD Disaggregation]`, `[Executor]`, `[Graph Optimization]`, `[Speculative Decoding]`, `[RL]`, `[Models]`, `[Quantization]`, `[Loader]`, `[OP]`, `[KVCache]`, `[DataProcessor]`, `[BugFix]`, `[Docs]`, `[CI]`, `[Optimization]`, `[Feature]`, `[Benchmark]`, `[Others]`, `[XPU]`, `[HPU]`, `[GCU]`, `[DCU]`, `[Iluvatar]`, `[Metax]`]
  - You can add new tags based on the PR content, but the semantics must be clear.
- [x] Format your code, run `pre-commit` before commit.
- [ ] Add unit tests. Please write the reason in this PR if no unit tests.（当前为临时验证 PR，正式合入前补充 `tests/graph_optimization/` 单测）
- [ ] Provide accuracy results.
- [ ] If the current PR is submitting to the `release` branch, make sure the PR has been submitted to the `develop` branch, then cherry-pick it to the `release` branch with the `[Cherry-Pick]` PR tag.

总体评价

代码逻辑改动方向正确：在 deterministic 模式下，先重置 RNG 和 share_inputs 再执行 CUDAGraph capture，可确保捕获的图从确定性初始状态出发，修复思路清晰。建议正式合入前完善 PR 规范、补充多硬件同步检查，并将临时 [just test] 前缀去除。

[PaddlePaddle-bot]

🟡 建议 多硬件同步检查

gpu_worker.py 修改了 CUDAGraph capture 的执行顺序（capture_model() 移到 deterministic reset 之后），如果其他硬件 Worker（dcu_worker.py、xpu_worker.py、iluvatar_worker.py、metax_worker.py 等）中有相同的 CUDAGraph capture 逻辑，是否也存在同样的 deterministic 模式问题？

建议：检查 fastdeploy/worker/{xpu,dcu,gcu,hpu,iluvatar,metax}_worker.py 中是否有类似 capture 顺序问题，如有，同步修复。